Junq's Arxiv
Computation and Language
☆ Rewarding Chatbots for Real-World Engagement with Millions of Users
The emergence of pretrained large language models has led to the deployment of a range of social chatbots for chitchat. Although these chatbots demonstrate language ability and fluency, they are not guaranteed to be engaging and can struggle to retain users. This work investigates the development of social chatbots that prioritize user engagement to enhance retention, specifically examining the use of human feedback to efficiently develop highly engaging chatbots. The proposed approach uses automatic pseudo-labels collected from user interactions to train a reward model that can be used to reject low-scoring sample responses generated by the chatbot model at inference time. Intuitive evaluation metrics, such as mean conversation length (MCL), are introduced as proxies to measure the level of engagement of deployed chatbots. A/B testing on groups of 10,000 new daily chatbot users on the Chai Research platform shows that this approach increases the MCL by up to 70%, which translates to a more than 30% increase in user retention for a GPT-J 6B model. Future work aims to use the reward model to realise a data fly-wheel, where the latest user conversations can be used to alternately fine-tune the language model and the reward model.
☆ Susceptibility to Influence of Large Language Models
Two studies tested the hypothesis that a Large Language Model (LLM) can be used to model psychological change following exposure to influential input. The first study tested a generic mode of influence - the Illusory Truth Effect (ITE) - where earlier exposure to a statement (through, for example, rating its interest) boosts a later truthfulness test rating. Data was collected from 1000 human participants using an online experiment, and 1000 simulated participants using engineered prompts and LLM completion. 64 ratings per participant were collected, using all exposure-test combinations of the attributes: truth, interest, sentiment and importance. The results for human participants reconfirmed the ITE, and demonstrated an absence of effect for attributes other than truth, and when the same attribute is used for exposure and test. The same pattern of effects was found for LLM-simulated participants. The second study concerns a specific mode of influence - populist framing of news to increase its persuasion and political mobilization. Data from LLM-simulated participants was collected and compared to previously published data from a 15-country experiment on 7286 human participants. Several effects previously demonstrated from the human study were replicated by the simulated study, including effects that surprised the authors of the human study by contradicting their theoretical expectations (anti-immigrant framing of news decreases its persuasion and mobilization); but some significant relationships found in human data (modulation of the effectiveness of populist framing according to relative deprivation of the participant) were not present in the LLM data. Together the two studies support the view that LLMs have potential to act as models of the effect of influence.
comment: 24 pages, 6 figures, 7 tables, 53 references
☆ Is In-hospital Meta-information Useful for Abstractive Discharge Summary Generation?
During the patient's hospitalization, the physician must record daily observations of the patient and summarize them into a brief document called "discharge summary" when the patient is discharged. Automated generation of discharge summary can greatly relieve the physicians' burden, and has been addressed recently in the research community. Most previous studies of discharge summary generation using the sequence-to-sequence architecture focus on only inpatient notes for input. However, electric health records (EHR) also have rich structured metadata (e.g., hospital, physician, disease, length of stay, etc.) that might be useful. This paper investigates the effectiveness of medical meta-information for summarization tasks. We obtain four types of meta-information from the EHR systems and encode each meta-information into a sequence-to-sequence model. Using Japanese EHRs, meta-information encoded models increased ROUGE-1 by up to 4.45 points and BERTScore by 3.77 points over the vanilla Longformer. Also, we found that the encoded meta-information improves the precisions of its related terms in the outputs. Our results showed the benefit of the use of medical meta-information.
☆ Robust Knowledge Distillation from RNN-T Models With Noisy Training Labels Using Full-Sum Loss ICASSP 2023
This work studies knowledge distillation (KD) and addresses its constraints for recurrent neural network transducer (RNN-T) models. In hard distillation, a teacher model transcribes large amounts of unlabelled speech to train a student model. Soft distillation is another popular KD method that distills the output logits of the teacher model. Due to the nature of RNN-T alignments, applying soft distillation between RNN-T architectures having different posterior distributions is challenging. In addition, bad teachers having high word-error-rate (WER) reduce the efficacy of KD. We investigate how to effectively distill knowledge from variable quality ASR teachers, which has not been studied before to the best of our knowledge. We show that a sequence-level KD, full-sum distillation, outperforms other distillation methods for RNN-T models, especially for bad teachers. We also propose a variant of full-sum distillation that distills the sequence discriminative knowledge of the teacher leading to further improvement in WER. We conduct experiments on public datasets namely SpeechStew and LibriSpeech, and on in-house production data.
comment: Accepted at ICASSP 2023
☆ Creation and evaluation of timelines for longitudinal user posts EACL 2023
There is increasing interest to work with user generated content in social media, especially textual posts over time. Currently there is no consistent way of segmenting user posts into timelines in a meaningful way that improves the quality and cost of manual annotation. Here we propose a set of methods for segmenting longitudinal user posts into timelines likely to contain interesting moments of change in a user's behaviour, based on their online posting activity. We also propose a novel framework for evaluating timelines and show its applicability in the context of two different social media datasets. Finally, we present a discussion of the linguistic content of highly ranked timelines.
comment: Accepted at EACL 2023 (main, long); camera-ready version
☆ An algebraic approach to translating Japanese
We use Lambek's pregroups and the framework of compositional distributional models of language ("DisCoCat") to study translations from Japanese to English as pairs of functors. Adding decorations to pregroups we show how to handle word order changes between languages.
comment: 20 pages, multiple diagrams and glosses
☆ An Overview on Language Models: Recent Developments and Outlook
Language modeling studies the probability distributions over strings of texts. It is one of the most fundamental tasks in natural language processing (NLP). It has been widely used in text generation, speech recognition, machine translation, etc. Conventional language models (CLMs) aim to predict the probability of linguistic sequences in a causal manner. In contrast, pre-trained language models (PLMs) cover broader concepts and can be used in both causal sequential modeling and fine-tuning for downstream applications. PLMs have their own training paradigms (usually self-supervised) and serve as foundation models in modern NLP systems. This overview paper provides an introduction to both CLMs and PLMs from five aspects, i.e., linguistic units, structures, training methods, evaluation methods, and applications. Furthermore, we discuss the relationship between CLMs and PLMs and shed light on the future directions of language modeling in the pre-trained era.
☆ MIXPGD: Hybrid Adversarial Training for Speech Recognition Systems
Automatic speech recognition (ASR) systems based on deep neural networks are weak against adversarial perturbations. We propose mixPGD adversarial training method to improve the robustness of the model for ASR systems. In standard adversarial training, adversarial samples are generated by leveraging supervised or unsupervised methods. We merge the capabilities of both supervised and unsupervised approaches in our method to generate new adversarial samples which aid in improving model robustness. Extensive experiments and comparison across various state-of-the-art defense methods and adversarial attacks have been performed to show that mixPGD gains 4.1% WER of better performance than previous best performing models under white-box adversarial attack setting. We tested our proposed defense method against both white-box and transfer based black-box attack settings to ensure that our defense strategy is robust against various types of attacks. Empirical results on several adversarial attacks validate the effectiveness of our proposed approach.
☆ Clinical BERTScore: An Improved Measure of Automatic Speech Recognition Performance in Clinical Settings
Automatic Speech Recognition (ASR) in medical contexts has the potential to save time, cut costs, increase report accuracy, and reduce physician burnout. However, the healthcare industry has been slower to adopt this technology, in part due to the importance of avoiding medically-relevant transcription mistakes. In this work, we present the Clinical BERTScore (CBERTScore), an ASR metric that penalizes clinically-relevant mistakes more than others. We demonstrate that this metric more closely aligns with clinician preferences on medical sentences as compared to other metrics (WER, BLUE, METEOR, etc), sometimes by wide margins. We collect a benchmark of 13 clinician preferences on 149 realistic medical sentences called the Clinician Transcript Preference benchmark (CTP), demonstrate that CBERTScore more closely matches what clinicians prefer, and release the benchmark for the community to further develop clinically-aware ASR metrics.
☆ MuLTI: Efficient Video-and-Language Understanding with MultiWay-Sampler and Multiple Choice Modeling
Video-and-language understanding has a variety of applications in the industry, such as video question answering, text-video retrieval and multi-label classification. Existing video-and-language understanding methods generally adopt heavy multi-modal encoders and feature fusion modules, which consume large amounts of GPU memory. Especially, they have difficulty dealing with dense video frames or long text that are prevalent in industrial applications. In this paper, we propose MuLTI, a highly accurate and memory-efficient video-and-language understanding model that achieves efficient and effective feature fusion through feature sampling and attention modules. Therefore, MuLTI can handle longer sequences with limited GPU memory. Then, we introduce an attention-based adapter to the encoders, which finetunes the shallow features to improve the model's performance with low GPU memory consumption. Finally, to further improve the model's performance, we introduce a new pretraining task named Multiple Choice Modeling to bridge the task gap between pretraining and downstream tasks and enhance the model's ability to align the video and the text. Benefiting from the efficient feature fusion module, the attention-based adapter and the new pretraining task, MuLTI achieves state-of-the-art performance on multiple datasets. Implementation and pretrained models will be released.
☆ Logic Against Bias: Textual Entailment Mitigates Stereotypical Sentence Reasoning EACL 2023
Due to their similarity-based learning objectives, pretrained sentence encoders often internalize stereotypical assumptions that reflect the social biases that exist within their training corpora. In this paper, we describe several kinds of stereotypes concerning different communities that are present in popular sentence representation models, including pretrained next sentence prediction and contrastive sentence representation models. We compare such models to textual entailment models that learn language logic for a variety of downstream language understanding tasks. By comparing strong pretrained models based on text similarity with textual entailment learning, we conclude that the explicit logic learning with textual entailment can significantly reduce bias and improve the recognition of social communities, without an explicit de-biasing process
comment: Accepted by EACL 2023
♻ ☆ Arabic aspect sentiment polarity classification using BERT
Aspect-based sentiment analysis(ABSA) is a textual analysis methodology that defines the polarity of opinions on certain aspects related to specific targets. The majority of research on ABSA is in English, with a small amount of work available in Arabic. Most previous Arabic research has relied on deep learning models that depend primarily on context-independent word embeddings (e.g.word2vec), where each word has a fixed representation independent of its context. This article explores the modeling capabilities of contextual embeddings from pre-trained language models, such as BERT, and making use of sentence pair input on Arabic aspect sentiment polarity classification task. In particular, we develop a simple but effective BERT-based neural baseline to handle this task. Our BERT architecture with a simple linear classification layer surpassed the state-of-the-art works, according to the experimental results on three different Arabic datasets. Achieving an accuracy of 89.51% on the Arabic hotel reviews dataset, 73% on the Human annotated book reviews dataset, and 85.73% on the Arabic news dataset.
♻ ☆ GPT-3-driven pedagogical agents for training children's curious question-asking skills
In order to train children's ability to ask curiosity-driven questions, previous research has explored designing specific exercises relying on providing semantic and linguistic cues to help formulate such questions. But despite showing pedagogical efficiency, this method is still limited as it relies on generating the said cues by hand, which can be a very costly process. In this context, we propose to leverage advances in the natural language processing field (NLP) and investigate the efficiency of using a large language model (LLM) for automating the production of the pedagogical content of a curious question-asking (QA) training. We study generating the said content using the "prompt-based" method that consists of explaining the task to the LLM in natural text. We evaluate the output using human experts annotations and comparisons with hand-generated content. Results suggested indeed the relevance and usefulness of this content. We also conduct a field study in primary school (75 children aged 9-10), where we evaluate children's QA performance when having this training. We compare 3 types of content : 1) hand-generated content that proposes "closed" cues leading to predefined questions; 2) GPT-3-generated content that proposes the same type of cues; 3) GPT-3-generated content that proposes "open" cues leading to several possible questions. We see a similar QA performance between the two "closed" trainings (showing the scalability of the approach using GPT-3), and a better one for participants with the "open" training. These results suggest the efficiency of using LLMs to support children in generating more curious questions, using a natural language prompting approach that affords usability by teachers and other users not specialists of AI techniques. Furthermore, results also show that open-ended content may be more suitable for training curious question-asking skills.
♻ ☆ Large Language Models Are Human-Level Prompt Engineers
By conditioning on natural language instructions, large language models (LLMs) have displayed impressive capabilities as general-purpose computers. However, task performance depends significantly on the quality of the prompt used to steer the model, and most effective prompts have been handcrafted by humans. Inspired by classical program synthesis and the human approach to prompt engineering, we propose Automatic Prompt Engineer (APE) for automatic instruction generation and selection. In our method, we treat the instruction as the "program," optimized by searching over a pool of instruction candidates proposed by an LLM in order to maximize a chosen score function. To evaluate the quality of the selected instruction, we evaluate the zero-shot performance of another LLM following the selected instruction. Experiments on 24 NLP tasks show that our automatically generated instructions outperform the prior LLM baseline by a large margin and achieve better or comparable performance to the instructions generated by human annotators on 19/24 tasks. We conduct extensive qualitative and quantitative analyses to explore the performance of APE. We show that APE-engineered prompts can be applied to steer models toward truthfulness and/or informativeness, as well as to improve few-shot learning performance by simply prepending them to standard in-context learning prompts. Please check out our webpage at https://sites.google.com/view/automatic-prompt-engineer.
♻ ☆ A Kind Introduction to Lexical and Grammatical Aspect, with a Survey of Computational Approaches EACL 2023
Aspectual meaning refers to how the internal temporal structure of situations is presented. This includes whether a situation is described as a state or as an event, whether the situation is finished or ongoing, and whether it is viewed as a whole or with a focus on a particular phase. This survey gives an overview of computational approaches to modeling lexical and grammatical aspect along with intuitive explanations of the necessary linguistic concepts and terminology. In particular, we describe the concepts of stativity, telicity, habituality, perfective and imperfective, as well as influential inventories of eventuality and situation types. We argue that because aspect is a crucial component of semantics, especially when it comes to reporting the temporal structure of situations in a precise way, future NLP approaches need to be able to handle and evaluate it systematically in order to achieve human-level language understanding.
comment: Accepted at EACL 2023, camera ready version
♻ ☆ Self-Adaptive Named Entity Recognition by Retrieving Unstructured Knowledge EACL2023
Although named entity recognition (NER) helps us to extract domain-specific entities from text (e.g., artists in the music domain), it is costly to create a large amount of training data or a structured knowledge base to perform accurate NER in the target domain. Here, we propose self-adaptive NER, which retrieves external knowledge from unstructured text to learn the usages of entities that have not been learned well. To retrieve useful knowledge for NER, we design an effective two-stage model that retrieves unstructured knowledge using uncertain entities as queries. Our model predicts the entities in the input and then finds those of which the prediction is not confident. Then, it retrieves knowledge by using these uncertain entities as queries and concatenates the retrieved text to the original input to revise the prediction. Experiments on CrossNER datasets demonstrated that our model outperforms strong baselines by 2.35 points in F1 metric.
comment: EACL2023 (long)
♻ ☆ Topic Modelling of Swedish Newspaper Articles about Coronavirus: a Case Study using Latent Dirichlet Allocation Method
Topic Modelling (TM) is from the research branches of natural language understanding (NLU) and natural language processing (NLP) that is to facilitate insightful analysis from large documents and datasets, such as a summarisation of main topics and the topic changes. This kind of discovery is getting more popular in real-life applications due to its impact on big data analytics. In this study, from the social-media and healthcare domain, we apply popular Latent Dirichlet Allocation (LDA) methods to model the topic changes in Swedish newspaper articles about Coronavirus. We describe the corpus we created including 6515 articles, methods applied, and statistics on topic changes over approximately 1 year and two months period of time from 17th January 2020 to 13th March 2021. We hope this work can be an asset for grounding applications of topic modelling and can be inspiring for similar case studies in an era with pandemics, to support socio-economic impact research as well as clinical and healthcare analytics. Our data and source code are openly available at https://github. com/poethan/Swed_Covid_TM Keywords: Latent Dirichlet Allocation (LDA); Topic Modelling; Coronavirus; Pandemics; Natural Language Understanding; BERT-topic
comment: 14 pages, 14 figures
♻ ☆ Fillers in Spoken Language Understanding: Computational and Psycholinguistic Perspectives
Disfluencies (i.e. interruptions in the regular flow of speech), are ubiquitous to spoken discourse. Fillers ("uh", "um") are disfluencies that occur the most frequently compared to other kinds of disfluencies. Yet, to the best of our knowledge, there isn't a resource that brings together the research perspectives influencing Spoken Language Understanding (SLU) on these speech events. This aim of this article is to survey a breadth of perspectives in a holistic way; i.e. from considering underlying (psycho)linguistic theory, to their annotation and consideration in Automatic Speech Recognition (ASR) and SLU systems, to lastly, their study from a generation standpoint. This article aims to present the perspectives in an approachable way to the SLU and Conversational AI community, and discuss moving forward, what we believe are the trends and challenges in each area.
comment: To appear in TAL Journal
♻ ☆ Word-Graph2vec: An efficient word embedding approach on word co-occurrence graph using random walk sampling
Word embedding has become ubiquitous and is widely used in various text mining and natural language processing (NLP) tasks, such as information retrieval, semantic analysis, and machine translation, among many others. Unfortunately, it is prohibitively expensive to train the word embedding in a relatively large corpus. We propose a graph-based word embedding algorithm, called Word-Graph2vec, which converts the large corpus into a word co-occurrence graph, then takes the word sequence samples from this graph by randomly traveling and trains the word embedding on this sampling corpus in the end. We posit that because of the stable vocabulary, relative idioms, and fixed expressions in English, the size and density of the word co-occurrence graph change slightly with the increase in the training corpus. So that Word-Graph2vec has stable runtime on the large scale data set, and its performance advantage becomes more and more obvious with the growth of the training corpus. Extensive experiments conducted on real-world datasets show that the proposed algorithm outperforms traditional Skip-Gram by four-five times in terms of efficiency, while the error generated by the random walk sampling is small.
♻ ☆ BERT-Deep CNN: State-of-the-Art for Sentiment Analysis of COVID-19 Tweets
The free flow of information has been accelerated by the rapid development of social media technology. There has been a significant social and psychological impact on the population due to the outbreak of Coronavirus disease (COVID-19). The COVID-19 pandemic is one of the current events being discussed on social media platforms. In order to safeguard societies from this pandemic, studying people's emotions on social media is crucial. As a result of their particular characteristics, sentiment analysis of texts like tweets remains challenging. Sentiment analysis is a powerful text analysis tool. It automatically detects and analyzes opinions and emotions from unstructured data. Texts from a wide range of sources are examined by a sentiment analysis tool, which extracts meaning from them, including emails, surveys, reviews, social media posts, and web articles. To evaluate sentiments, natural language processing (NLP) and machine learning techniques are used, which assign weights to entities, topics, themes, and categories in sentences or phrases. Machine learning tools learn how to detect sentiment without human intervention by examining examples of emotions in text. In a pandemic situation, analyzing social media texts to uncover sentimental trends can be very helpful in gaining a better understanding of society's needs and predicting future trends. We intend to study society's perception of the COVID-19 pandemic through social media using state-of-the-art BERT and Deep CNN models. The superiority of BERT models over other deep models in sentiment analysis is evident and can be concluded from the comparison of the various research studies mentioned in this article.
comment: 20 pages, 5 figures
♻ ☆ Temporal Modeling Matters: A Novel Temporal Emotional Modeling Approach for Speech Emotion Recognition ICASSP 2023
Speech emotion recognition (SER) plays a vital role in improving the interactions between humans and machines by inferring human emotion and affective states from speech signals. Whereas recent works primarily focus on mining spatiotemporal information from hand-crafted features, we explore how to model the temporal patterns of speech emotions from dynamic temporal scales. Towards that goal, we introduce a novel temporal emotional modeling approach for SER, termed Temporal-aware bI-direction Multi-scale Network (TIM-Net), which learns multi-scale contextual affective representations from various time scales. Specifically, TIM-Net first employs temporal-aware blocks to learn temporal affective representation, then integrates complementary information from the past and the future to enrich contextual representations, and finally, fuses multiple time scale features for better adaptation to the emotional variation. Extensive experimental results on six benchmark SER datasets demonstrate the superior performance of TIM-Net, gaining 2.34% and 2.61% improvements of the average UAR and WAR over the second-best on each corpus. The source code is available at https://github.com/Jiaxin-Ye/TIM-Net_SER.
comment: Accepted by ICASSP 2023
♻ ☆ CoSyn: Detecting Implicit Hate Speech in Online Conversations Using a Context Synergized Hyperbolic Network IJCAI 2023
The tremendous growth of social media users interacting in online conversations has also led to significant growth in hate speech. Most of the prior works focus on detecting explicit hate speech, which is overt and leverages hateful phrases, with very little work focusing on detecting hate speech that is implicit or denotes hatred through indirect or coded language. In this paper, we present CoSyn, a user- and conversational-context synergized network for detecting implicit hate speech in online conversation trees. CoSyn first models the user's personal historical and social context using a novel hyperbolic Fourier attention mechanism and hyperbolic graph convolution network. Next, we jointly model the user's personal context and the conversational context using a novel context interaction mechanism in the hyperbolic space that clearly captures the interplay between the two and makes independent assessments on the amounts of information to be retrieved from both contexts. CoSyn performs all operations in the hyperbolic space to account for the scale-free dynamics of social media. We demonstrate the effectiveness of CoSyn both qualitatively and quantitatively on an open-source hate speech dataset with Twitter conversations and show that CoSyn outperforms all our baselines in detecting implicit hate speech with absolute improvements in the range of 8.15% - 19.50%.
comment: Under review at IJCAI 2023
♻ ☆ Self-Attention Networks Can Process Bounded Hierarchical Languages ACL 2021
Despite their impressive performance in NLP, self-attention networks were recently proved to be limited for processing formal languages with hierarchical structure, such as $\mathsf{Dyck}_k$, the language consisting of well-nested parentheses of $k$ types. This suggested that natural language can be approximated well with models that are too weak for formal languages, or that the role of hierarchy and recursion in natural language might be limited. We qualify this implication by proving that self-attention networks can process $\mathsf{Dyck}_{k, D}$, the subset of $\mathsf{Dyck}_{k}$ with depth bounded by $D$, which arguably better captures the bounded hierarchical structure of natural language. Specifically, we construct a hard-attention network with $D+1$ layers and $O(\log k)$ memory size (per token per layer) that recognizes $\mathsf{Dyck}_{k, D}$, and a soft-attention network with two layers and $O(\log k)$ memory size that generates $\mathsf{Dyck}_{k, D}$. Experiments show that self-attention networks trained on $\mathsf{Dyck}_{k, D}$ generalize to longer inputs with near-perfect accuracy, and also verify the theoretical memory advantage of self-attention networks over recurrent networks.
comment: ACL 2021. 19 pages with extended appendix. v2 fixed a small typo in the formula at the end of page 5 (thank to Gabriel Faria). Code: https://github.com/princeton-nlp/dyck-transformer
♻ ☆ ReAct: Synergizing Reasoning and Acting in Language Models ICLR
While large language models (LLMs) have demonstrated impressive capabilities across tasks in language understanding and interactive decision making, their abilities for reasoning (e.g. chain-of-thought prompting) and acting (e.g. action plan generation) have primarily been studied as separate topics. In this paper, we explore the use of LLMs to generate both reasoning traces and task-specific actions in an interleaved manner, allowing for greater synergy between the two: reasoning traces help the model induce, track, and update action plans as well as handle exceptions, while actions allow it to interface with external sources, such as knowledge bases or environments, to gather additional information. We apply our approach, named ReAct, to a diverse set of language and decision making tasks and demonstrate its effectiveness over state-of-the-art baselines, as well as improved human interpretability and trustworthiness over methods without reasoning or acting components. Concretely, on question answering (HotpotQA) and fact verification (Fever), ReAct overcomes issues of hallucination and error propagation prevalent in chain-of-thought reasoning by interacting with a simple Wikipedia API, and generates human-like task-solving trajectories that are more interpretable than baselines without reasoning traces. On two interactive decision making benchmarks (ALFWorld and WebShop), ReAct outperforms imitation and reinforcement learning methods by an absolute success rate of 34% and 10% respectively, while being prompted with only one or two in-context examples. Project site with code: https://react-lm.github.io
comment: v3 is the ICLR camera ready version with some typos fixed. Project site with code: https://react-lm.github.io
Computer Vision and Pattern Recognition
☆ StyleGANEX: StyleGAN-Based Manipulation Beyond Cropped Aligned Faces
Recent advances in face manipulation using StyleGAN have produced impressive results. However, StyleGAN is inherently limited to cropped aligned faces at a fixed image resolution it is pre-trained on. In this paper, we propose a simple and effective solution to this limitation by using dilated convolutions to rescale the receptive fields of shallow layers in StyleGAN, without altering any model parameters. This allows fixed-size small features at shallow layers to be extended into larger ones that can accommodate variable resolutions, making them more robust in characterizing unaligned faces. To enable real face inversion and manipulation, we introduce a corresponding encoder that provides the first-layer feature of the extended StyleGAN in addition to the latent style code. We validate the effectiveness of our method using unaligned face inputs of various resolutions in a diverse set of face manipulation tasks, including facial attribute editing, super-resolution, sketch/mask-to-face translation, and face toonification.
comment: Code: https://github.com/williamyang1991/StyleGANEX Project page: https://www.mmlab-ntu.com/project/styleganex/
☆ Learning to Select Camera Views: Efficient Multiview Understanding at Few Glances
Multiview camera setups have proven useful in many computer vision applications for reducing ambiguities, mitigating occlusions, and increasing field-of-view coverage. However, the high computational cost associated with multiple views poses a significant challenge for end devices with limited computational resources. To address this issue, we propose a view selection approach that analyzes the target object or scenario from given views and selects the next best view for processing. Our approach features a reinforcement learning based camera selection module, MVSelect, that not only selects views but also facilitates joint training with the task network. Experimental results on multiview classification and detection tasks show that our approach achieves promising performance while using only 2 or 3 out of N available views, significantly reducing computational costs. Furthermore, analysis on the selected views reveals that certain cameras can be shut off with minimal performance impact, shedding light on future camera layout optimization for multiview systems. Code is available at https://github.com/hou-yz/MVSelect.
☆ Learning Object-Centric Neural Scattering Functions for Free-viewpoint Relighting and Scene Composition
Photorealistic object appearance modeling from 2D images is a constant topic in vision and graphics. While neural implicit methods (such as Neural Radiance Fields) have shown high-fidelity view synthesis results, they cannot relight the captured objects. More recent neural inverse rendering approaches have enabled object relighting, but they represent surface properties as simple BRDFs, and therefore cannot handle translucent objects. We propose Object-Centric Neural Scattering Functions (OSFs) for learning to reconstruct object appearance from only images. OSFs not only support free-viewpoint object relighting, but also can model both opaque and translucent objects. While accurately modeling subsurface light transport for translucent objects can be highly complex and even intractable for neural methods, OSFs learn to approximate the radiance transfer from a distant light to an outgoing direction at any spatial location. This approximation avoids explicitly modeling complex subsurface scattering, making learning a neural implicit model tractable. Experiments on real and synthetic data show that OSFs accurately reconstruct appearances for both opaque and translucent objects, allowing faithful free-viewpoint relighting as well as scene composition.
comment: Journal extension of arXiv:2012.08503. The first two authors contributed equally to this work
☆ Single-branch Network for Multimodal Training ICASSP 2023
With the rapid growth of social media platforms, users are sharing billions of multimedia posts containing audio, images, and text. Researchers have focused on building autonomous systems capable of processing such multimedia data to solve challenging multimodal tasks including cross-modal retrieval, matching, and verification. Existing works use separate networks to extract embeddings of each modality to bridge the gap between them. The modular structure of their branched networks is fundamental in creating numerous multimodal applications and has become a defacto standard to handle multiple modalities. In contrast, we propose a novel single-branch network capable of learning discriminative representation of unimodal as well as multimodal tasks without changing the network. An important feature of our single-branch network is that it can be trained either using single or multiple modalities without sacrificing performance. We evaluated our proposed single-branch network on the challenging multimodal problem (face-voice association) for cross-modal verification and matching tasks with various loss formulations. Experimental results demonstrate the superiority of our proposed single-branch network over the existing methods in a wide range of experiments. Code: https://github.com/msaadsaeed/SBNet
comment: Accepted at ICASSP 2023
Self-supervised Training Sample Difficulty Balancing for Local Descriptor Learning
In the case of an imbalance between positive and negative samples, hard negative mining strategies have been shown to help models learn more subtle differences between positive and negative samples, thus improving recognition performance. However, if too strict mining strategies are promoted in the dataset, there may be a risk of introducing false negative samples. Meanwhile, the implementation of the mining strategy disrupts the difficulty distribution of samples in the real dataset, which may cause the model to over-fit these difficult samples. Therefore, in this paper, we investigate how to trade off the difficulty of the mined samples in order to obtain and exploit high-quality negative samples, and try to solve the problem in terms of both the loss function and the training strategy. The proposed balance loss provides an effective discriminant for the quality of negative samples by combining a self-supervised approach to the loss function, and uses a dynamic gradient modulation strategy to achieve finer gradient adjustment for samples of different difficulties. The proposed annealing training strategy then constrains the difficulty of the samples drawn from negative sample mining to provide data sources with different difficulty distributions for the loss function, and uses samples of decreasing difficulty to train the model. Extensive experiments show that our new descriptors outperform previous state-of-the-art descriptors for patch validation, matching, and retrieval tasks.
☆ Improving Domain-Invariance in Self-Supervised Learning via Batch Styles Standardization
The recent rise of Self-Supervised Learning (SSL) as one of the preferred strategies for learning with limited labeled data, and abundant unlabeled data has led to the widespread use of these models. They are usually pretrained, finetuned, and evaluated on the same data distribution, i.e., within an in-distribution setting. However, they tend to perform poorly in out-of-distribution evaluation scenarios, a challenge that Unsupervised Domain Generalization (UDG) seeks to address. This paper introduces a novel method to standardize the styles of images in a batch. Batch styles standardization, relying on Fourier-based augmentations, promotes domain invariance in SSL by preventing spurious correlations from leaking into the features. The combination of batch styles standardization with the well-known contrastive-based method SimCLR leads to a novel UDG method named CLaSSy ($\textbf{C}$ontrastive $\textbf{L}$e$\textbf{a}$rning with $\textbf{S}$tandardized $\textbf{S}$t$\textbf{y}$les). CLaSSy offers serious advantages over prior methods, as it does not rely on domain labels and is scalable to handle a large number of domains. Experimental results on various UDG datasets demonstrate the superior performance of CLaSSy compared to existing UDG methods. Finally, the versatility of the proposed batch styles standardization is demonstrated by extending respectively the contrastive-based and non-contrastive-based SSL methods, SWaV and MSN, while considering different backbone architectures (convolutional-based, transformers-based).
☆ Communication-Critical Planning via Multi-Agent Trajectory Exchange ICRA 2023
This paper addresses the task of joint multi-agent perception and planning, especially as it relates to the real-world challenge of collision-free navigation for connected self-driving vehicles. For this task, several communication-enabled vehicles must navigate through a busy intersection while avoiding collisions with each other and with obstacles. To this end, this paper proposes a learnable costmap-based planning mechanism, given raw perceptual data, that is (1) distributed, (2) uncertainty-aware, and (3) bandwidth-efficient. Our method produces a costmap and uncertainty-aware entropy map to sort and fuse candidate trajectories as evaluated across multiple-agents. The proposed method demonstrates several favorable performance trends on a suite of open-source overhead datasets as well as within a novel communication-critical simulator. It produces accurate semantic occupancy forecasts as an intermediate perception output, attaining a 72.5% average pixel-wise classification accuracy. By selecting the top trajectory, the multi-agent method scales well with the number of agents, reducing the hard collision rate by up to 57% with eight agents compared to the single-agent version.
comment: Accepted to ICRA 2023
☆ Long-tailed Classification from a Bayesian-decision-theory Perspective
Long-tailed classification poses a challenge due to its heavy imbalance in class probabilities and tail-sensitivity risks with asymmetric misprediction costs. Recent attempts have used re-balancing loss and ensemble methods, but they are largely heuristic and depend heavily on empirical results, lacking theoretical explanation. Furthermore, existing methods overlook the decision loss, which characterizes different costs associated with tailed classes. This paper presents a general and principled framework from a Bayesian-decision-theory perspective, which unifies existing techniques including re-balancing and ensemble methods, and provides theoretical justifications for their effectiveness. From this perspective, we derive a novel objective based on the integrated risk and a Bayesian deep-ensemble approach to improve the accuracy of all classes, especially the ``tail". Besides, our framework allows for task-adaptive decision loss which provides provably optimal decisions in varying task scenarios, along with the capability to quantify uncertainty. Finally, We conduct comprehensive experiments, including standard classification, tail-sensitive classification with a new False Head Rate metric, calibration, and ablation studies. Our framework significantly improves the current SOTA even on large-scale real-world datasets like ImageNet.
☆ MVImgNet: A Large-scale Dataset of Multi-view Images CVPR2023
Being data-driven is one of the most iconic properties of deep learning algorithms. The birth of ImageNet drives a remarkable trend of "learning from large-scale data" in computer vision. Pretraining on ImageNet to obtain rich universal representations has been manifested to benefit various 2D visual tasks, and becomes a standard in 2D vision. However, due to the laborious collection of real-world 3D data, there is yet no generic dataset serving as a counterpart of ImageNet in 3D vision, thus how such a dataset can impact the 3D community is unraveled. To remedy this defect, we introduce MVImgNet, a large-scale dataset of multi-view images, which is highly convenient to gain by shooting videos of real-world objects in human daily life. It contains 6.5 million frames from 219,188 videos crossing objects from 238 classes, with rich annotations of object masks, camera parameters, and point clouds. The multi-view attribute endows our dataset with 3D-aware signals, making it a soft bridge between 2D and 3D vision. We conduct pilot studies for probing the potential of MVImgNet on a variety of 3D and 2D visual tasks, including radiance field reconstruction, multi-view stereo, and view-consistent image understanding, where MVImgNet demonstrates promising performance, remaining lots of possibilities for future explorations. Besides, via dense reconstruction on MVImgNet, a 3D object point cloud dataset is derived, called MVPNet, covering 87,200 samples from 150 categories, with the class label on each point cloud. Experiments show that MVPNet can benefit the real-world 3D object classification while posing new challenges to point cloud understanding. MVImgNet and MVPNet will be publicly available, hoping to inspire the broader vision community.
comment: To be appear in CVPR2023. Project page: https://gaplab.cuhk.edu.cn/projects/MVImgNet/
☆ Importance of Aligning Training Strategy with Evaluation for Diffusion Models in 3D Multiclass Segmentation
Recently, denoising diffusion probabilistic models (DDPM) have been applied to image segmentation by generating segmentation masks conditioned on images, while the applications were mainly limited to 2D networks without exploiting potential benefits from the 3D formulation. In this work, for the first time, DDPMs are used for 3D multiclass image segmentation. We make three key contributions that all focus on aligning the training strategy with the evaluation methodology, and improving efficiency. Firstly, the model predicts segmentation masks instead of sampled noise and is optimised directly via Dice loss. Secondly, the predicted mask in the previous time step is recycled to generate noise-corrupted masks to reduce information leakage. Finally, the diffusion process during training was reduced to five steps, the same as the evaluation. Through studies on two large multiclass data sets (prostate MR and abdominal CT), we demonstrated significantly improved performance compared to existing DDPMs, and reached competitive performance with non-diffusion segmentation models, based on U-net, within the same compute budget. The JAX-based diffusion framework has been released on https://github.com/mathpluscode/ImgX-DiffSeg.
☆ Dynamic Y-KD: A Hybrid Approach to Continual Instance Segmentation
Despite the success of deep learning methods on instance segmentation, these models still suffer from catastrophic forgetting in continual learning scenarios. In this paper, our contributions for continual instance segmentation are threefold. First, we propose the Y-knowledge distillation (Y-KD), a knowledge distillation strategy that shares a common feature extractor between the teacher and student networks. As the teacher is also updated with new data in Y-KD, the increased plasticity results in new modules that are specialized on new classes. Second, our Y-KD approach is supported by a dynamic architecture method that grows new modules for each task and uses all of them for inference with a unique instance segmentation head, which significantly reduces forgetting. Third, we complete our approach by leveraging checkpoint averaging as a simple method to manually balance the trade-off between the performance on the various sets of classes, thus increasing the control over the model's behavior without any additional cost. These contributions are united in our model that we name the Dynamic Y-KD network. We perform extensive experiments on several single-step and multi-steps scenarios on Pascal-VOC, and we show that our approach outperforms previous methods both on past and new classes. For instance, compared to recent work, our method obtains +2.1% mAP on old classes in 15-1, +7.6% mAP on new classes in 19-1 and reaches 91.5% of the mAP obtained by joint-training on all classes in 15-5.
☆ Combining visibility analysis and deep learning for refinement of semantic 3D building models by conflict classification
Semantic 3D building models are widely available and used in numerous applications. Such 3D building models display rich semantics but no fa\c{c}ade openings, chiefly owing to their aerial acquisition techniques. Hence, refining models' fa\c{c}ades using dense, street-level, terrestrial point clouds seems a promising strategy. In this paper, we propose a method of combining visibility analysis and neural networks for enriching 3D models with window and door features. In the method, occupancy voxels are fused with classified point clouds, which provides semantics to voxels. Voxels are also used to identify conflicts between laser observations and 3D models. The semantic voxels and conflicts are combined in a Bayesian network to classify and delineate fa\c{c}ade openings, which are reconstructed using a 3D model library. Unaffected building semantics is preserved while the updated one is added, thereby upgrading the building model to LoD3. Moreover, Bayesian network results are back-projected onto point clouds to improve points' classification accuracy. We tested our method on a municipal CityGML LoD2 repository and the open point cloud datasets: TUM-MLS-2016 and TUM-FA\c{C}ADE. Validation results revealed that the method improves the accuracy of point cloud semantic segmentation and upgrades buildings with fa\c{c}ade elements. The method can be applied to enhance the accuracy of urban simulations and facilitate the development of semantic segmentation algorithms.
comment: ISPRS Annals, 3DGeoInfo 2022, Australia, Sydney
☆ New Benchmarks for Accountable Text-based Visual Re-creation
Given a command, humans can directly execute the action after thinking or choose to reject it, with reasonable feedback at the same time. However, the behavior of existing text-to-image generation methods are uncontrollable and irresponsible. In this paper, we construct extensive experiments to verify whether they can be accountable (say no and explain why) for those prohibited instructions. To this end, we define a novel text-based visual re-creation task and construct new synthetic CLEVR-NOT dataset (620K) and manually pictured Fruit-NOT dataset (50K). In our method, one text-image pair as the query is fed into the machine, and the model gives a yes or no answer after visual and textual reasoning. If the answer is yes, the image auto-encoder and auto-regressive transformer must complete the visual re-creation under the premise of ensuring image quality, otherwise the system needs to explain why the commands cannot be completed or prohibited. We provide a detailed analysis of experimental results in image quality, answer accuracy, and model behavior in the face of uncertainty and imperfect user queries. Our results demonstrate the difficulty of a single model for both textual and visual reasoning. We also hope our explorations and findings can bring valuable insights about the accountability of text-based image generation models. Code and datasets can be found at https://matrix-alpha.github.io.
comment: 13 pages, 9 figures
☆ Open-Ended Medical Visual Question Answering Through Prefix Tuning of Language Models
Medical Visual Question Answering (VQA) is an important challenge, as it would lead to faster and more accurate diagnoses and treatment decisions. Most existing methods approach it as a multi-class classification problem, which restricts the outcome to a predefined closed-set of curated answers. We focus on open-ended VQA and motivated by the recent advances in language models consider it as a generative task. Leveraging pre-trained language models, we introduce a novel method particularly suited for small, domain-specific, medical datasets. To properly communicate the medical images to the language model, we develop a network that maps the extracted visual features to a set of learnable tokens. Then, alongside the question, these learnable tokens directly prompt the language model. We explore recent parameter-efficient fine-tuning strategies for language models, which allow for resource- and data-efficient fine-tuning. We evaluate our approach on the prime medical VQA benchmarks, namely, Slake, OVQA and PathVQA. The results demonstrate that our approach outperforms existing methods across various training settings while also being computationally efficient.
☆ Exploring Recurrent Long-term Temporal Fusion for Multi-view 3D Perception
Long-term temporal fusion is a crucial but often overlooked technique in camera-based Bird's-Eye-View (BEV) 3D perception. Existing methods are mostly in a parallel manner. While parallel fusion can benefit from long-term information, it suffers from increasing computational and memory overheads as the fusion window size grows. Alternatively, BEVFormer adopts a recurrent fusion pipeline so that history information can be efficiently integrated, yet it fails to benefit from longer temporal frames. In this paper, we explore an embarrassingly simple long-term recurrent fusion strategy built upon the LSS-based methods and find it already able to enjoy the merits from both sides, i.e., rich long-term information and efficient fusion pipeline. A temporal embedding module is further proposed to improve the model's robustness against occasionally missed frames in practical scenarios. We name this simple but effective fusing pipeline VideoBEV. Experimental results on the nuScenes benchmark show that VideoBEV obtains leading performance on various camera-based 3D perception tasks, including object detection (55.4% mAP and 62.9% NDS), segmentation (48.6% vehicle mIoU), tracking (54.8% AMOTA), and motion prediction (0.80m minADE and 0.463 EPA). Code will be available.
☆ Score-Based Generative Models for Medical Image Segmentation using Signed Distance Functions
Medical image segmentation is a crucial task that relies on the ability to accurately identify and isolate regions of interest in images. Thereby, generative approaches allow to capture the statistical properties of segmentation masks that are dependent on the respective medical images. In this work we propose a conditional score-based generative modeling framework that leverages the signed distance function to represent an implicit and smoother distribution of segmentation masks. The score function of the conditional distribution of segmentation masks is learned in a conditional denoising process, which can be effectively used to generate accurate segmentation masks. Moreover, uncertainty maps can be generated, which can aid in further analysis and thus enhance the predictive robustness. We qualitatively and quantitatively illustrate competitive performance of the proposed method on a public nuclei and gland segmentation data set, highlighting its potential utility in medical image segmentation applications.
☆ Automated crack propagation measurement on asphalt concrete specimens using an optical flow-based deep neural network
This article proposes a deep neural network, namely CrackPropNet, to measure crack propagation on asphalt concrete (AC) specimens. It offers an accurate, flexible, efficient, and low-cost solution for crack propagation measurement using images collected during cracking tests. CrackPropNet significantly differs from traditional deep learning networks, as it involves learning to locate displacement field discontinuities by matching features at various locations in the reference and deformed images. An image library representing the diversified cracking behavior of AC was developed for supervised training. CrackPropNet achieved an optimal dataset scale F-1 of 0.755 and optimal image scale F-1 of 0.781 on the testing dataset at a running speed of 26 frame-per-second. Experiments demonstrated that low to medium-level Gaussian noises had a limited impact on the measurement accuracy of CrackPropNet. Moreover, the model showed promising generalization on fundamentally different images. As a crack measurement technique, the CrackPropNet can detect complex crack patterns accurately and efficiently in AC cracking tests. It can be applied to characterize the cracking phenomenon, evaluate AC cracking potential, validate test protocols, and verify theoretical models.
☆ Neuron Structure Modeling for Generalizable Remote Physiological Measurement CVPR2023
Remote photoplethysmography (rPPG) technology has drawn increasing attention in recent years. It can extract Blood Volume Pulse (BVP) from facial videos, making many applications like health monitoring and emotional analysis more accessible. However, as the BVP signal is easily affected by environmental changes, existing methods struggle to generalize well for unseen domains. In this paper, we systematically address the domain shift problem in the rPPG measurement task. We show that most domain generalization methods do not work well in this problem, as domain labels are ambiguous in complicated environmental changes. In light of this, we propose a domain-label-free approach called NEuron STructure modeling (NEST). NEST improves the generalization capacity by maximizing the coverage of feature space during training, which reduces the chance for under-optimized feature activation during inference. Besides, NEST can also enrich and enhance domain invariant features across multi-domain. We create and benchmark a large-scale domain generalization protocol for the rPPG measurement task. Extensive experiments show that our approach outperforms the state-of-the-art methods on both cross-dataset and intra-dataset settings.
comment: Accepted by CVPR2023
☆ Understanding and Constructing Latent Modality Structures in Multi-modal Representation Learning CVPR 2023
Contrastive loss has been increasingly used in learning representations from multiple modalities. In the limit, the nature of the contrastive loss encourages modalities to exactly match each other in the latent space. Yet it remains an open question how the modality alignment affects the downstream task performance. In this paper, based on an information-theoretic argument, we first prove that exact modality alignment is sub-optimal in general for downstream prediction tasks. Hence we advocate that the key of better performance lies in meaningful latent modality structures instead of perfect modality alignment. To this end, we propose three general approaches to construct latent modality structures. Specifically, we design 1) a deep feature separation loss for intra-modality regularization; 2) a Brownian-bridge loss for inter-modality regularization; and 3) a geometric consistency loss for both intra- and inter-modality regularization. Extensive experiments are conducted on two popular multi-modal representation learning frameworks: the CLIP-based two-tower model and the ALBEF-based fusion model. We test our model on a variety of tasks including zero/few-shot image classification, image-text retrieval, visual question answering, visual reasoning, and visual entailment. Our method achieves consistent improvements over existing methods, demonstrating the effectiveness and generalizability of our proposed approach on latent modality structure regularization.
comment: 14 pages, 8 figure, CVPR 2023 accepted
☆ ACR: Attention Collaboration-based Regressor for Arbitrary Two-Hand Reconstruction CVPR 2023
Reconstructing two hands from monocular RGB images is challenging due to frequent occlusion and mutual confusion. Existing methods mainly learn an entangled representation to encode two interacting hands, which are incredibly fragile to impaired interaction, such as truncated hands, separate hands, or external occlusion. This paper presents ACR (Attention Collaboration-based Regressor), which makes the first attempt to reconstruct hands in arbitrary scenarios. To achieve this, ACR explicitly mitigates interdependencies between hands and between parts by leveraging center and part-based attention for feature extraction. However, reducing interdependence helps release the input constraint while weakening the mutual reasoning about reconstructing the interacting hands. Thus, based on center attention, ACR also learns cross-hand prior that handle the interacting hands better. We evaluate our method on various types of hand reconstruction datasets. Our method significantly outperforms the best interacting-hand approaches on the InterHand2.6M dataset while yielding comparable performance with the state-of-the-art single-hand methods on the FreiHand dataset. More qualitative results on in-the-wild and hand-object interaction datasets and web images/videos further demonstrate the effectiveness of our approach for arbitrary hand reconstruction. Our code is available at https://github.com/ZhengdiYu/Arbitrary-Hands-3D-Reconstruction.
comment: Accepted by CVPR 2023; Code at https://github.com/ZhengdiYu/Arbitrary-Hands-3D-Reconstruction
☆ Structural Multiplane Image: Bridging Neural View Synthesis and 3D Reconstruction CVPR2023
The Multiplane Image (MPI), containing a set of fronto-parallel RGBA layers, is an effective and efficient representation for view synthesis from sparse inputs. Yet, its fixed structure limits the performance, especially for surfaces imaged at oblique angles. We introduce the Structural MPI (S-MPI), where the plane structure approximates 3D scenes concisely. Conveying RGBA contexts with geometrically-faithful structures, the S-MPI directly bridges view synthesis and 3D reconstruction. It can not only overcome the critical limitations of MPI, i.e., discretization artifacts from sloped surfaces and abuse of redundant layers, and can also acquire planar 3D reconstruction. Despite the intuition and demand of applying S-MPI, great challenges are introduced, e.g., high-fidelity approximation for both RGBA layers and plane poses, multi-view consistency, non-planar regions modeling, and efficient rendering with intersected planes. Accordingly, we propose a transformer-based network based on a segmentation model. It predicts compact and expressive S-MPI layers with their corresponding masks, poses, and RGBA contexts. Non-planar regions are inclusively handled as a special case in our unified framework. Multi-view consistency is ensured by sharing global proxy embeddings, which encode plane-level features covering the complete 3D scenes with aligned coordinates. Intensive experiments show that our method outperforms both previous state-of-the-art MPI-based view synthesis methods and planar reconstruction methods.
comment: Accepted to CVPR2023
☆ Boosting Open-Set Domain Adaptation with Threshold Self-Tuning and Cross-Domain Mixup
Open-set domain adaptation (OSDA) aims to not only recognize target samples belonging to common classes shared by source and target domains but also perceive unknown class samples. Existing OSDA methods suffer from two obstacles. First, a tedious process of manually tuning a hyperparameter $threshold$ is required for most OSDA approaches to separate common and unknown classes. It is difficult to determine a proper threshold when the target domain data is unlabeled. Second, most OSDA methods only rely on confidence values predicted by models to distinguish common/unknown classes. The performance is not satisfied, especially when the majority of the target domain consists of unknown classes. Our experiments demonstrate that combining entropy, consistency, and confidence is a more reliable way of distinguishing common and unknown samples. In this paper, we design a novel threshold self-tuning and cross-domain mixup (TSCM) method to overcome the two drawbacks. TSCM can automatically tune a proper threshold utilizing unlabeled target samples rather than manually setting an empirical hyperparameter. Our method considers multiple criteria instead of only the confidence and uses the threshold generated by itself to separate common and unknown classes in the target domain. Furthermore, we introduce a cross-domain mixup method designed for OSDA scenarios to learn domain-invariant features in a more continuous latent space. Comprehensive experiments illustrate that our method consistently achieves superior performance on different benchmarks compared with various state-of-the-arts.
☆ Marginalia and machine learning: Handwritten text recognition for Marginalia Collections
The pressing need for digitization of historical document collections has led to a strong interest in designing computerised image processing methods for automatic handwritten text recognition (HTR). Handwritten text possesses high variability due to different writing styles, languages and scripts. Training an accurate and robust HTR system calls for data-efficient approaches due to the unavailability of sufficient amounts of annotated multi-writer text. A case study on an ongoing project ``Marginalia and Machine Learning" is presented here that focuses on automatic detection and recognition of handwritten marginalia texts i.e., text written in margins or handwritten notes. Faster R-CNN network is used for detection of marginalia and AttentionHTR is used for word recognition. The data comes from early book collections (printed) found in the Uppsala University Library, with handwritten marginalia texts. Source code and pretrained models are available at https://github.com/ektavats/Project-Marginalia.
comment: Work under progress
☆ Estimating friction coefficient using generative modelling
It is common to utilise dynamic models to measure the tyre-road friction in real-time. Alternatively, predictive approaches estimate the tyre-road friction by identifying the environmental factors affecting it. This work aims to formulate the problem of friction estimation as a visual perceptual learning task. The problem is broken down into detecting surface characteristics by applying semantic segmentation and using the extracted features to predict the frictional force. This work for the first time formulates the friction estimation problem as a regression from the latent space of a semantic segmentation model. The preliminary results indicate that this approach can estimate frictional force.
comment: To be published in ICM2023
☆ GECCO: Geometrically-Conditioned Point Diffusion Models
Diffusion models generating images conditionally on text, such as Dall-E 2 and Stable Diffusion, have recently made a splash far beyond the computer vision community. Here, we tackle the related problem of generating point clouds, both unconditionally, and conditionally with images. For the latter, we introduce a novel geometrically-motivated conditioning scheme based on projecting sparse image features into the point cloud and attaching them to each individual point, at every step in the denoising process. This approach improves geometric consistency and yields greater fidelity than current methods relying on unstructured, global latent codes. Additionally, we show how to apply recent continuous-time diffusion schemes. Our method performs on par or above the state of art on conditional and unconditional experiments on synthetic data, while being faster, lighter, and delivering tractable likelihoods. We show it can also scale to diverse indoors scenes.
☆ DACov: A Deeper Analysis of Data Augmentation on the Computed Tomography Segmentation Problem
Due to the COVID-19 global pandemic, computer-assisted diagnoses of medical images have gained much attention, and robust methods of semantic segmentation of Computed Tomography (CT) images have become highly desirable. In this work, we present a deeper analysis of how data augmentation techniques improve segmentation performance on this problem. We evaluate 20 traditional augmentation techniques on five public datasets. Six different probabilities of applying each augmentation technique on an image were evaluated. We also assess a different training methodology where the training subsets are combined into a single larger set. All networks were evaluated through a 5-fold cross-validation strategy, resulting in over 4,600 experiments. We also propose a novel data augmentation technique based on Generative Adversarial Networks (GANs) to create new healthy and unhealthy lung CT images, evaluating four variations of our approach with the same six probabilities of the traditional methods. Our findings show that GAN-based techniques and spatial-level transformations are the most promising for improving the learning of deep models on this problem, with the StarGANv2 + F with a probability of 0.3 achieving the highest F-score value on the Ricord1a dataset in the unified training strategy. Our code is publicly available at https://github.com/VRI-UFPR/DACov2022
☆ Object-Aware Distillation Pyramid for Open-Vocabulary Object Detection CVPR 2023
Open-vocabulary object detection aims to provide object detectors trained on a fixed set of object categories with the generalizability to detect objects described by arbitrary text queries. Previous methods adopt knowledge distillation to extract knowledge from Pretrained Vision-and-Language Models (PVLMs) and transfer it to detectors. However, due to the non-adaptive proposal cropping and single-level feature mimicking processes, they suffer from information destruction during knowledge extraction and inefficient knowledge transfer. To remedy these limitations, we propose an Object-Aware Distillation Pyramid (OADP) framework, including an Object-Aware Knowledge Extraction (OAKE) module and a Distillation Pyramid (DP) mechanism. When extracting object knowledge from PVLMs, the former adaptively transforms object proposals and adopts object-aware mask attention to obtain precise and complete knowledge of objects. The latter introduces global and block distillation for more comprehensive knowledge transfer to compensate for the missing relation information in object distillation. Extensive experiments show that our method achieves significant improvement compared to current methods. Especially on the MS-COCO dataset, our OADP framework reaches $35.6$ mAP$^{\text{N}}_{50}$, surpassing the current state-of-the-art method by $3.3$ mAP$^{\text{N}}_{50}$. Code is released at https://github.com/LutingWang/OADP.
comment: Accepted by CVPR 2023
☆ Bi3D: Bi-domain Active Learning for Cross-domain 3D Object Detection CVPR2023
Unsupervised Domain Adaptation (UDA) technique has been explored in 3D cross-domain tasks recently. Though preliminary progress has been made, the performance gap between the UDA-based 3D model and the supervised one trained with fully annotated target domain is still large. This motivates us to consider selecting partial-yet-important target data and labeling them at a minimum cost, to achieve a good trade-off between high performance and low annotation cost. To this end, we propose a Bi-domain active learning approach, namely Bi3D, to solve the cross-domain 3D object detection task. The Bi3D first develops a domainness-aware source sampling strategy, which identifies target-domain-like samples from the source domain to avoid the model being interfered by irrelevant source data. Then a diversity-based target sampling strategy is developed, which selects the most informative subset of target domain to improve the model adaptability to the target domain using as little annotation budget as possible. Experiments are conducted on typical cross-domain adaptation scenarios including cross-LiDAR-beam, cross-country, and cross-sensor, where Bi3D achieves a promising target-domain detection accuracy (89.63% on KITTI) compared with UDAbased work (84.29%), even surpassing the detector trained on the full set of the labeled target domain (88.98%). Our code is available at: https://github.com/PJLabADG/3DTrans.
comment: Accepted by CVPR2023; Code is available at https://github.com/PJLabADG/3DTrans
☆ Handheld Burst Super-Resolution Meets Multi-Exposure Satellite Imagery
Image resolution is an important criterion for many applications based on satellite imagery. In this work, we adapt a state-of-the-art kernel regression technique for smartphone camera burst super-resolution to satellites. This technique leverages the local structure of the image to optimally steer the fusion kernels, limiting blur in the final high-resolution prediction, denoising the image, and recovering details up to a zoom factor of 2. We extend this approach to the multi-exposure case to predict from a sequence of multi-exposure low-resolution frames a high-resolution and noise-free one. Experiments on both single and multi-exposure scenarios show the merits of the approach. Since the fusion is learning-free, the proposed method is ensured to not hallucinate details, which is crucial for many remote sensing applications.
comment: 9 pages
☆ Accurate Real-time Polyp Detection in Videos from Concatenation of Latent Features Extracted from Consecutive Frames
An efficient deep learning model that can be implemented in real-time for polyp detection is crucial to reducing polyp miss-rate during screening procedures. Convolutional neural networks (CNNs) are vulnerable to small changes in the input image. A CNN-based model may miss the same polyp appearing in a series of consecutive frames and produce unsubtle detection output due to changes in camera pose, lighting condition, light reflection, etc. In this study, we attempt to tackle this problem by integrating temporal information among neighboring frames. We propose an efficient feature concatenation method for a CNN-based encoder-decoder model without adding complexity to the model. The proposed method incorporates extracted feature maps of previous frames to detect polyps in the current frame. The experimental results demonstrate that the proposed method of feature concatenation improves the overall performance of automatic polyp detection in videos. The following results are obtained on a public video dataset: sensitivity 90.94\%, precision 90.53\%, and specificity 92.46%
☆ You Only Train Once: Multi-Identity Free-Viewpoint Neural Human Rendering from Monocular Videos
We introduce You Only Train Once (YOTO), a dynamic human generation framework, which performs free-viewpoint rendering of different human identities with distinct motions, via only one-time training from monocular videos. Most prior works for the task require individualized optimization for each input video that contains a distinct human identity, leading to a significant amount of time and resources for the deployment, thereby impeding the scalability and the overall application potential of the system. In this paper, we tackle this problem by proposing a set of learnable identity codes to expand the capability of the framework for multi-identity free-viewpoint rendering, and an effective pose-conditioned code query mechanism to finely model the pose-dependent non-rigid motions. YOTO optimizes neural radiance fields (NeRF) by utilizing designed identity codes to condition the model for learning various canonical T-pose appearances in a single shared volumetric representation. Besides, our joint learning of multiple identities within a unified model incidentally enables flexible motion transfer in high-quality photo-realistic renderings for all learned appearances. This capability expands its potential use in important applications, including Virtual Reality. We present extensive experimental results on ZJU-MoCap and PeopleSnapshot to clearly demonstrate the effectiveness of our proposed model. YOTO shows state-of-the-art performance on all evaluation metrics while showing significant benefits in training and inference efficiency as well as rendering quality. The code and model will be made publicly available soon.
☆ Contrastive Language-Image Pretrained (CLIP) Models are Powerful Out-of-Distribution Detectors
We present a comprehensive experimental study on pretrained feature extractors for visual out-of-distribution (OOD) detection. We examine several setups, based on the availability of labels or image captions and using different combinations of in- and out-distributions. Intriguingly, we find that (i) contrastive language-image pretrained models achieve state-of-the-art unsupervised out-of-distribution performance using nearest neighbors feature similarity as the OOD detection score, (ii) supervised state-of-the-art OOD detection performance can be obtained without in-distribution fine-tuning, (iii) even top-performing billion-scale vision transformers trained with natural language supervision fail at detecting adversarially manipulated OOD images. Finally, we argue whether new benchmarks for visual anomaly detection are needed based on our experiments. Using the largest publicly available vision transformer, we achieve state-of-the-art performance across all $18$ reported OOD benchmarks, including an AUROC of 87.6\% (9.2\% gain, unsupervised) and 97.4\% (1.2\% gain, supervised) for the challenging task of CIFAR100 $\rightarrow$ CIFAR10 OOD detection. The code will be open-sourced.
☆ Aleth-NeRF: Low-light Condition View Synthesis with Concealing Fields
Common capture low-light scenes are challenging for most computer vision techniques, including Neural Radiance Fields (NeRF). Vanilla NeRF is viewer-centred that simplifies the rendering process only as light emission from 3D locations in the viewing direction, thus failing to model the low-illumination induced darkness. Inspired by emission theory of ancient Greek that visual perception is accomplished by rays casting from eyes, we make slight modifications on vanilla NeRF to train on multiple views of low-light scene, we can thus render out the well-lit scene in an unsupervised manner. We introduce a surrogate concept, Concealing Fields, that reduce the transport of light during the volume rendering stage. Specifically, our proposed method, Aleth-NeRF, directly learns from the dark image to understand volumetric object representation and concealing field under priors. By simply eliminating Concealing Fields, we can render a single or multi-view well-lit image(s) and gain superior performance over other 2D low light enhancement methods. Additionally, we collect the first paired LOw-light and normal-light Multi-view (LOM) datasets for future research.
comment: website page: https://cuiziteng.github.io/Aleth_NeRF_web/
☆ Enhancing the success rates by performing pooling decisions adjacent to the output layer
Learning classification tasks of (2^nx2^n) inputs typically consist of \le n (2x2) max-pooling (MP) operators along the entire feedforward deep architecture. Here we show, using the CIFAR-10 database, that pooling decisions adjacent to the last convolutional layer significantly enhance accuracy success rates (SRs). In particular, average SRs of the advanced VGG with m layers (A-VGGm) architectures are 0.936, 0.940, 0.954, 0.955, and 0.955 for m=6, 8, 14, 13, and 16, respectively. The results indicate A-VGG8s' SR is superior to VGG16s', and that the SRs of A-VGG13 and A-VGG16 are equal, and comparable to that of Wide-ResNet16. In addition, replacing the three fully connected (FC) layers with one FC layer, A-VGG6 and A-VGG14, or with several linear activation FC layers, yielded similar SRs. These significantly enhanced SRs stem from training the most influential input-output routes, in comparison to the inferior routes selected following multiple MP decisions along the deep architecture. In addition, SRs are sensitive to the order of the non-commutative MP and average pooling operators adjacent to the output layer, varying the number and location of training routes. The results call for the reexamination of previously proposed deep architectures and their SRs by utilizing the proposed pooling strategy adjacent to the output layer.
comment: 27 pages, 3 figures, 1 table and Supplementary Information
☆ AnoMalNet: Outlier Detection based Malaria Cell Image Classification Method Leveraging Deep Autoencoder
Class imbalance is a pervasive issue in the field of disease classification from medical images. It is necessary to balance out the class distribution while training a model for decent results. However, in the case of rare medical diseases, images from affected patients are much harder to come by compared to images from non-affected patients, resulting in unwanted class imbalance. Various processes of tackling class imbalance issues have been explored so far, each having its fair share of drawbacks. In this research, we propose an outlier detection based binary medical image classification technique which can handle even the most extreme case of class imbalance. We have utilized a dataset of malaria parasitized and uninfected cells. An autoencoder model titled AnoMalNet is trained with only the uninfected cell images at the beginning and then used to classify both the affected and non-affected cell images by thresholding a loss value. We have achieved an accuracy, precision, recall, and F1 score of 98.49%, 97.07%, 100%, and 98.52% respectively, performing better than large deep learning models and other published works. As our proposed approach can provide competitive results without needing the disease-positive samples during training, it should prove to be useful in binary disease classification on imbalanced datasets.
☆ Scaling Up 3D Kernels with Bayesian Frequency Re-parameterization for Medical Image Segmentation
With the inspiration of vision transformers, the concept of depth-wise convolution revisits to provide a large Effective Receptive Field (ERF) using Large Kernel (LK) sizes for medical image segmentation. However, the segmentation performance might be saturated and even degraded as the kernel sizes scaled up (e.g., $21\times 21\times 21$) in a Convolutional Neural Network (CNN). We hypothesize that convolution with LK sizes is limited to maintain an optimal convergence for locality learning. While Structural Re-parameterization (SR) enhances the local convergence with small kernels in parallel, optimal small kernel branches may hinder the computational efficiency for training. In this work, we propose RepUX-Net, a pure CNN architecture with a simple large kernel block design, which competes favorably with current network state-of-the-art (SOTA) (e.g., 3D UX-Net, SwinUNETR) using 6 challenging public datasets. We derive an equivalency between kernel re-parameterization and the branch-wise variation in kernel convergence. Inspired by the spatial frequency in the human visual system, we extend to vary the kernel convergence into element-wise setting and model the spatial frequency as a Bayesian prior to re-parameterize convolutional weights during training. Specifically, a reciprocal function is leveraged to estimate a frequency-weighted value, which rescales the corresponding kernel element for stochastic gradient descent. From the experimental results, RepUX-Net consistently outperforms 3D SOTA benchmarks with internal validation (FLARE: 0.929 to 0.944), external validation (MSD: 0.901 to 0.932, KiTS: 0.815 to 0.847, LiTS: 0.933 to 0.949, TCIA: 0.736 to 0.779) and transfer learning (AMOS: 0.880 to 0.911) scenarios in Dice Score.
comment: Both codes and pretrained models are available at: https://github.com/MASILab/RepUX-Net
☆ Knowledge Transfer via Multi-Head Feature Adaptation for Whole Slide Image Classification
Transferring prior knowledge from a source domain to the same or similar target domain can greatly enhance the performance of models on the target domain. However, it is challenging to directly leverage the knowledge from the source domain due to task discrepancy and domain shift. To bridge the gaps between different tasks and domains, we propose a Multi-Head Feature Adaptation module, which projects features in the source feature space to a new space that is more similar to the target space. Knowledge transfer is particularly important in Whole Slide Image (WSI) classification since the number of WSIs in one dataset might be too small to achieve satisfactory performance. Therefore, WSI classification is an ideal testbed for our method, and we adapt multiple knowledge transfer methods for WSI classification. The experimental results show that models with knowledge transfer outperform models that are trained from scratch by a large margin regardless of the number of WSIs in the datasets, and our method achieves state-of-the-art performances among other knowledge transfer methods on multiple datasets, including TCGA-RCC, TCGA-NSCLC, and Camelyon16 datasets.
Self-Supervised CSF Inpainting with Synthetic Atrophy for Improved Accuracy Validation of Cortical Surface Analyses
Accuracy validation of cortical thickness measurement is a difficult problem due to the lack of ground truth data. To address this need, many methods have been developed to synthetically induce gray matter (GM) atrophy in an MRI via deformable registration, creating a set of images with known changes in cortical thickness. However, these methods often cause blurring in atrophied regions, and cannot simulate realistic atrophy within deep sulci where cerebrospinal fluid (CSF) is obscured or absent. In this paper, we present a solution using a self-supervised inpainting model to generate CSF in these regions and create images with more plausible GM/CSF boundaries. Specifically, we introduce a novel, 3D GAN model that incorporates patch-based dropout training, edge map priors, and sinusoidal positional encoding, all of which are established methods previously limited to 2D domains. We show that our framework significantly improves the quality of the resulting synthetic images and is adaptable to unseen data with fine-tuning. We also demonstrate that our resulting dataset can be employed for accuracy validation of cortical segmentation and thickness measurement.
comment: Accepted at Medical Imaging with Deep Learning (MIDL) 2023
☆ Self-NeRF: A Self-Training Pipeline for Few-Shot Neural Radiance Fields
Recently, Neural Radiance Fields (NeRF) have emerged as a potent method for synthesizing novel views from a dense set of images. Despite its impressive performance, NeRF is plagued by its necessity for numerous calibrated views and its accuracy diminishes significantly in a few-shot setting. To address this challenge, we propose Self-NeRF, a self-evolved NeRF that iteratively refines the radiance fields with very few number of input views, without incorporating additional priors. Basically, we train our model under the supervision of reference and unseen views simultaneously in an iterative procedure. In each iteration, we label unseen views with the predicted colors or warped pixels generated by the model from the preceding iteration. However, these expanded pseudo-views are afflicted by imprecision in color and warping artifacts, which degrades the performance of NeRF. To alleviate this issue, we construct an uncertainty-aware NeRF with specialized embeddings. Some techniques such as cone entropy regularization are further utilized to leverage the pseudo-views in the most efficient manner. Through experiments under various settings, we verified that our Self-NeRF is robust to input with uncertainty and surpasses existing methods when trained on limited training data.
comment: 11 pages, 11 figures
☆ Learning Global-Local Correspondence with Semantic Bottleneck for Logical Anomaly Detection
This paper presents a novel framework, named Global-Local Correspondence Framework (GLCF), for visual anomaly detection with logical constraints. Visual anomaly detection has become an active research area in various real-world applications, such as industrial anomaly detection and medical disease diagnosis. However, most existing methods focus on identifying local structural degeneration anomalies and often fail to detect high-level functional anomalies that involve logical constraints. To address this issue, we propose a two-branch approach that consists of a local branch for detecting structural anomalies and a global branch for detecting logical anomalies. To facilitate local-global feature correspondence, we introduce a novel semantic bottleneck enabled by the visual Transformer. Moreover, we develop feature estimation networks for each branch separately to detect anomalies. Our proposed framework is validated using various benchmarks, including industrial datasets, Mvtec AD, Mvtec Loco AD, and the Retinal-OCT medical dataset. Experimental results show that our method outperforms existing methods, particularly in detecting logical anomalies.
comment: Submission to IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
☆ Automatic Detection and Rectification of Paper Receipts on Smartphones
We describe the development of a real-time smartphone app that allows the user to digitize paper receipts in a novel way by "waving" their phone over the receipts and letting the app automatically detect and rectify the receipts for subsequent text recognition. We show that traditional computer vision algorithms for edge and corner detection do not robustly detect the non-linear and discontinuous edges and corners of a typical paper receipt in real-world settings. This is particularly the case when the colors of the receipt and background are similar, or where other interfering rectangular objects are present. Inaccurate detection of a receipt's corner positions then results in distorted images when using an affine projective transformation to rectify the perspective. We propose an innovative solution to receipt corner detection by treating each of the four corners as a unique "object", and training a Single Shot Detection MobileNet object detection model. We use a small amount of real data and a large amount of automatically generated synthetic data that is designed to be similar to real-world imaging scenarios. We show that our proposed method robustly detects the four corners of a receipt, giving a receipt detection accuracy of 85.3% on real-world data, compared to only 36.9% with a traditional edge detection-based approach. Our method works even when the color of the receipt is virtually indistinguishable from the background. Moreover, our method is trained to detect only the corners of the central target receipt and implicitly learns to ignore other receipts, and other rectangular objects. Including synthetic data allows us to train an even better model. These factors are a major advantage over traditional edge detection-based approaches, allowing us to deliver a much better experience to the user.
☆ TrojDiff: Trojan Attacks on Diffusion Models with Diverse Targets CVPR2023
Diffusion models have achieved great success in a range of tasks, such as image synthesis and molecule design. As such successes hinge on large-scale training data collected from diverse sources, the trustworthiness of these collected data is hard to control or audit. In this work, we aim to explore the vulnerabilities of diffusion models under potential training data manipulations and try to answer: How hard is it to perform Trojan attacks on well-trained diffusion models? What are the adversarial targets that such Trojan attacks can achieve? To answer these questions, we propose an effective Trojan attack against diffusion models, TrojDiff, which optimizes the Trojan diffusion and generative processes during training. In particular, we design novel transitions during the Trojan diffusion process to diffuse adversarial targets into a biased Gaussian distribution and propose a new parameterization of the Trojan generative process that leads to an effective training objective for the attack. In addition, we consider three types of adversarial targets: the Trojaned diffusion models will always output instances belonging to a certain class from the in-domain distribution (In-D2D attack), out-of-domain distribution (Out-D2D-attack), and one specific instance (D2I attack). We evaluate TrojDiff on CIFAR-10 and CelebA datasets against both DDPM and DDIM diffusion models. We show that TrojDiff always achieves high attack performance under different adversarial targets using different types of triggers, while the performance in benign environments is preserved. The code is available at https://github.com/chenweixin107/TrojDiff.
comment: CVPR2023
☆ Fast Diffusion Sampler for Inverse Problems by Geometric Decomposition
Diffusion models have shown exceptional performance in solving inverse problems. However, one major limitation is the slow inference time. While faster diffusion samplers have been developed for unconditional sampling, there has been limited research on conditional sampling in the context of inverse problems. In this study, we propose a novel and efficient diffusion sampling strategy that employs the geometric decomposition of diffusion sampling. Specifically, we discover that the samples generated from diffusion models can be decomposed into two orthogonal components: a ``denoised" component obtained by projecting the sample onto the clean data manifold, and a ``noise" component that induces a transition to the next lower-level noisy manifold with the addition of stochastic noise. Furthermore, we prove that, under some conditions on the clean data manifold, the conjugate gradient update for imposing conditioning from the denoised signal belongs to the clean manifold, resulting in a much faster and more accurate diffusion sampling. Our method is applicable regardless of the parameterization and setting (i.e., VE, VP). Notably, we achieve state-of-the-art reconstruction quality on challenging real-world medical inverse imaging problems, including multi-coil MRI reconstruction and 3D CT reconstruction. Moreover, our proposed method achieves more than 80 times faster inference time than the previous state-of-the-art method.
comment: 21 pages
☆ Deep Learning for Predicting Metastasis on Melanoma WSIs
Northern Europe has the second highest mortality rate of melanoma globally. In 2020, the mortality rate of melanoma rose to 1.9 per 100 000 habitants. Melanoma prognosis is based on a pathologist's subjective visual analysis of the patient's tumor. This methodology is heavily time-consuming, and the prognosis variability among experts is notable, drastically jeopardizing its reproducibility. Thus, the need for faster and more reproducible methods arises. Machine learning has paved its way into digital pathology, but so far, most contributions are on localization, segmentation, and diagnostics, with little emphasis on prognostics. This paper presents a convolutional neural network (CNN) method based on VGG16 to predict melanoma prognosis as the presence of metastasis within five years. Patches are extracted from regions of interest from Whole Slide Images (WSIs) at different magnification levels used in model training and validation. Results infer that utilizing WSI patches at 20x magnification level has the best performance, with an F1 score of 0.7667 and an AUC of 0.81.
☆ Multi-site, Multi-domain Airway Tree Modeling (ATM'22): A Public Benchmark for Pulmonary Airway Segmentation
Open international challenges are becoming the de facto standard for assessing computer vision and image analysis algorithms. In recent years, new methods have extended the reach of pulmonary airway segmentation that is closer to the limit of image resolution. Since EXACT'09 pulmonary airway segmentation, limited effort has been directed to quantitative comparison of newly emerged algorithms driven by the maturity of deep learning based approaches and clinical drive for resolving finer details of distal airways for early intervention of pulmonary diseases. Thus far, public annotated datasets are extremely limited, hindering the development of data-driven methods and detailed performance evaluation of new algorithms. To provide a benchmark for the medical imaging community, we organized the Multi-site, Multi-domain Airway Tree Modeling (ATM'22), which was held as an official challenge event during the MICCAI 2022 conference. ATM'22 provides large-scale CT scans with detailed pulmonary airway annotation, including 500 CT scans (300 for training, 50 for validation, and 150 for testing). The dataset was collected from different sites and it further included a portion of noisy COVID-19 CTs with ground-glass opacity and consolidation. Twenty-three teams participated in the entire phase of the challenge and the algorithms for the top ten teams are reviewed in this paper. Quantitative and qualitative results revealed that deep learning models embedded with the topological continuity enhancement achieved superior performance in general. ATM'22 challenge holds as an open-call design, the training data and the gold standard evaluation are available upon successful registration via its homepage.
comment: 32 pages, 16 figures. Homepage: https://atm22.grand-challenge.org/. Submitted
☆ Boosting Semi-Supervised Few-Shot Object Detection with SoftER Teacher
Few-shot object detection is an emerging problem aimed at detecting novel concepts from few exemplars. Existing approaches to few-shot detection assume abundant base labels to adapt to novel objects. This paper explores the task of semi-supervised few-shot detection by considering a realistic scenario which lacks abundant labels for both base and novel objects. Motivated by this unique problem, we introduce SoftER Teacher, a robust detector combining the advantages of pseudo-labeling with representation learning on region proposals. SoftER Teacher harnesses unlabeled data to jointly optimize for semi-supervised few-shot detection without explicitly relying on abundant base labels. Extensive experiments show that SoftER Teacher matches the novel class performance of a strong supervised detector using only 10% of base labels. Our work also sheds insight into a previously unknown relationship between semi-supervised and few-shot detection to suggest that a stronger semi-supervised detector leads to a more label-efficient few-shot detector. Code and models are available at https://github.com/lexisnexis-risk-open-source/ledetection
comment: Technical Report
☆ Generative Model Based Noise Robust Training for Unsupervised Domain Adaptation
Target domain pseudo-labelling has shown effectiveness in unsupervised domain adaptation (UDA). However, pseudo-labels of unlabeled target domain data are inevitably noisy due to the distribution shift between source and target domains. This paper proposes a Generative model-based Noise-Robust Training method (GeNRT), which eliminates domain shift while mitigating label noise. GeNRT incorporates a Distribution-based Class-wise Feature Augmentation (D-CFA) and a Generative-Discriminative classifier Consistency (GDC), both based on the class-wise target distributions modelled by generative models. D-CFA minimizes the domain gap by augmenting the source data with distribution-sampled target features, and trains a noise-robust discriminative classifier by using target domain knowledge from the generative models. GDC regards all the class-wise generative models as generative classifiers and enforces a consistency regularization between the generative and discriminative classifiers. It exploits an ensemble of target knowledge from all the generative models to train a noise-robust discriminative classifier and eventually gets theoretically linked to the Ben-David domain adaptation theorem for reducing the domain gap. Extensive experiments on Office-Home, PACS, and Digit-Five show that our GeNRT achieves comparable performance to state-of-the-art methods under single-source and multi-source UDA settings.
☆ IC classifier: a classifier for 3D industrial components based on geometric prior using GNN
In this paper, we propose an approach to address the problem of classifying 3D industrial components by introducing a novel framework named IC-classifier (Industrial Component classifier). Our framework is designed to focus on the object's local and global structures, emphasizing the former by incorporating specific local features for embedding the model. By utilizing graphical neural networks and embedding derived from geometric properties, IC-classifier facilitates the exploration of the local structures of the object while using geometric attention for the analysis of global structures. Furthermore, the framework uses point clouds to circumvent the heavy computation workload. The proposed framework's performance is benchmarked against state-of-the-art models, demonstrating its potential to compete in the field.
comment: 15 pages including citations, 3 pages of figures
☆ CVT-SLR: Contrastive Visual-Textual Transformation for Sign Language Recognition with Variational Alignment
Sign language recognition (SLR) is a weakly supervised task that annotates sign videos as textual glosses. Recent studies show that insufficient training caused by the lack of large-scale available sign language datasets becomes the main bottleneck for SLR. The majority of SLR works thereby adopt pretrained visual modules and develop two mainstream solutions. The multi-stream architectures extend multi-cue visual features, yielding the current SOTA performances but requiring complex designs and might introduce potential noise. Alternatively, the advanced single-cue SLR frameworks using explicit cross-modal alignment between visual and textual modalities are simple and effective, potentially competitive with the multi-cue framework. In this work, we propose a novel contrastive visual-textual transformation for SLR, CVT-SLR, to fully explore the pretrained knowledge of both the visual and language modalities. Based on the single-cue cross-modal alignment framework, we propose a variational autoencoder (VAE) for pretrained contextual knowledge while introducing the complete pretrained language module. The VAE implicitly aligns visual and textual modalities while benefiting from pretrained contextual knowledge as the traditional contextual module. Meanwhile, a contrastive cross-modal alignment algorithm is proposed to further enhance the explicit consistency constraints. Extensive experiments conducted on the two most popular public datasets, PHOENIX-2014 and PHOENIX-2014T, demonstrate that our proposed SLR framework not only consistently outperforms existing single-cue methods but even outperforms SOTA multi-cue methods.
☆ 3D Cinemagraphy from a Single Image CVPR 2023
We present 3D Cinemagraphy, a new technique that marries 2D image animation with 3D photography. Given a single still image as input, our goal is to generate a video that contains both visual content animation and camera motion. We empirically find that naively combining existing 2D image animation and 3D photography methods leads to obvious artifacts or inconsistent animation. Our key insight is that representing and animating the scene in 3D space offers a natural solution to this task. To this end, we first convert the input image into feature-based layered depth images using predicted depth values, followed by unprojecting them to a feature point cloud. To animate the scene, we perform motion estimation and lift the 2D motion into the 3D scene flow. Finally, to resolve the problem of hole emergence as points move forward, we propose to bidirectionally displace the point cloud as per the scene flow and synthesize novel views by separately projecting them into target image planes and blending the results. Extensive experiments demonstrate the effectiveness of our method. A user study is also conducted to validate the compelling rendering results of our method.
comment: Accepted by CVPR 2023. Project page: https://xingyi-li.github.io/3d-cinemagraphy/
☆ Boosting Adversarial Attacks by Leveraging Decision Boundary Information
Due to the gap between a substitute model and a victim model, the gradient-based noise generated from a substitute model may have low transferability for a victim model since their gradients are different. Inspired by the fact that the decision boundaries of different models do not differ much, we conduct experiments and discover that the gradients of different models are more similar on the decision boundary than in the original position. Moreover, since the decision boundary in the vicinity of an input image is flat along most directions, we conjecture that the boundary gradients can help find an effective direction to cross the decision boundary of the victim models. Based on it, we propose a Boundary Fitting Attack to improve transferability. Specifically, we introduce a method to obtain a set of boundary points and leverage the gradient information of these points to update the adversarial examples. Notably, our method can be combined with existing gradient-based methods. Extensive experiments prove the effectiveness of our method, i.e., improving the success rate by 5.6% against normally trained CNNs and 14.9% against defense CNNs on average compared to state-of-the-art transfer-based attacks. Further we compare transformers with CNNs, the results indicate that transformers are more robust than CNNs. However, our method still outperforms existing methods when attacking transformers. Specifically, when using CNNs as substitute models, our method obtains an average attack success rate of 58.2%, which is 10.8% higher than other state-of-the-art transfer-based attacks.
☆ Context-Based Trit-Plane Coding for Progressive Image Compression CVPR 2023
Trit-plane coding enables deep progressive image compression, but it cannot use autoregressive context models. In this paper, we propose the context-based trit-plane coding (CTC) algorithm to achieve progressive compression more compactly. First, we develop the context-based rate reduction module to estimate trit probabilities of latent elements accurately and thus encode the trit-planes compactly. Second, we develop the context-based distortion reduction module to refine partial latent tensors from the trit-planes and improve the reconstructed image quality. Third, we propose a retraining scheme for the decoder to attain better rate-distortion tradeoffs. Extensive experiments show that CTC outperforms the baseline trit-plane codec significantly in BD-rate on the Kodak lossless dataset, while increasing the time complexity only marginally. Our codes are available at https://github.com/seungminjeon-github/CTC.
comment: Accepted to CVPR 2023
Self-supervised Facial Action Unit Detection with Region and Relation Learning ICASSP 2023
Facial action unit (AU) detection is a challenging task due to the scarcity of manual annotations. Recent works on AU detection with self-supervised learning have emerged to address this problem, aiming to learn meaningful AU representations from numerous unlabeled data. However, most existing AU detection works with self-supervised learning utilize global facial features only, while AU-related properties such as locality and relevance are not fully explored. In this paper, we propose a novel self-supervised framework for AU detection with the region and relation learning. In particular, AU related attention map is utilized to guide the model to focus more on AU-specific regions to enhance the integrity of AU local features. Meanwhile, an improved Optimal Transport (OT) algorithm is introduced to exploit the correlation characteristics among AUs. In addition, Swin Transformer is exploited to model the long-distance dependencies within each AU region during feature learning. The evaluation results on BP4D and DISFA demonstrate that our proposed method is comparable or even superior to the state-of-the-art self-supervised learning methods and supervised AU detection methods.
comment: Accepted by ICASSP 2023
☆ MuLTI: Efficient Video-and-Language Understanding with MultiWay-Sampler and Multiple Choice Modeling
Video-and-language understanding has a variety of applications in the industry, such as video question answering, text-video retrieval and multi-label classification. Existing video-and-language understanding methods generally adopt heavy multi-modal encoders and feature fusion modules, which consume large amounts of GPU memory. Especially, they have difficulty dealing with dense video frames or long text that are prevalent in industrial applications. In this paper, we propose MuLTI, a highly accurate and memory-efficient video-and-language understanding model that achieves efficient and effective feature fusion through feature sampling and attention modules. Therefore, MuLTI can handle longer sequences with limited GPU memory. Then, we introduce an attention-based adapter to the encoders, which finetunes the shallow features to improve the model's performance with low GPU memory consumption. Finally, to further improve the model's performance, we introduce a new pretraining task named Multiple Choice Modeling to bridge the task gap between pretraining and downstream tasks and enhance the model's ability to align the video and the text. Benefiting from the efficient feature fusion module, the attention-based adapter and the new pretraining task, MuLTI achieves state-of-the-art performance on multiple datasets. Implementation and pretrained models will be released.
☆ MovingParts: Motion-based 3D Part Discovery in Dynamic Radiance Field
We present MovingParts, a NeRF-based method for dynamic scene reconstruction and part discovery. We consider motion as an important cue for identifying parts, that all particles on the same part share the common motion pattern. From the perspective of fluid simulation, existing deformation-based methods for dynamic NeRF can be seen as parameterizing the scene motion under the Eulerian view, i.e., focusing on specific locations in space through which the fluid flows as time passes. However, it is intractable to extract the motion of constituting objects or parts using the Eulerian view representation. In this work, we introduce the dual Lagrangian view and enforce representations under the Eulerian/Lagrangian views to be cycle-consistent. Under the Lagrangian view, we parameterize the scene motion by tracking the trajectory of particles on objects. The Lagrangian view makes it convenient to discover parts by factorizing the scene motion as a composition of part-level rigid motions. Experimentally, our method can achieve fast and high-quality dynamic scene reconstruction from even a single moving camera, and the induced part-based representation allows direct applications of part tracking, animation, 3D scene editing, etc.
comment: 10 pages
☆ Feature Unlearning for Generative Models via Implicit Feedback
We tackle the problem of feature unlearning from a pretrained image generative model. Unlike a common unlearning task where an unlearning target is a subset of the training set, we aim to unlearn a specific feature, such as hairstyle from facial images, from the pretrained generative models. As the target feature is only presented in a local region of an image, unlearning the entire image from the pretrained model may result in losing other details in the remaining region of the image. To specify which features to unlearn, we develop an implicit feedback mechanism where a user can select images containing the target feature. From the implicit feedback, we identify a latent representation corresponding to the target feature and then use the representation to unlearn the generative model. Our framework is generalizable for the two well-known families of generative models: GANs and VAEs. Through experiments on MNIST and CelebA datasets, we show that target features are successfully removed while keeping the fidelity of the original models.
☆ Explainable Semantic Medical Image Segmentation with Style
Semantic medical image segmentation using deep learning has recently achieved high accuracy, making it appealing to clinical problems such as radiation therapy. However, the lack of high-quality semantically labelled data remains a challenge leading to model brittleness to small shifts to input data. Most works require extra data for semi-supervised learning and lack the interpretability of the boundaries of the training data distribution during training, which is essential for model deployment in clinical practice. We propose a fully supervised generative framework that can achieve generalisable segmentation with only limited labelled data by simultaneously constructing an explorable manifold during training. The proposed approach creates medical image style paired with a segmentation task driven discriminator incorporating end-to-end adversarial training. The discriminator is generalised to small domain shifts as much as permissible by the training data, and the generator automatically diversifies the training samples using a manifold of input features learnt during segmentation. All the while, the discriminator guides the manifold learning by supervising the semantic content and fine-grained features separately during the image diversification. After training, visualisation of the learnt manifold from the generator is available to interpret the model limits. Experiments on a fully semantic, publicly available pelvis dataset demonstrated that our method is more generalisable to shifts than other state-of-the-art methods while being more explainable using an explorable manifold.
☆ Mode-locking Theory for Long-Range Interaction in Artificial Neural Networks
Visual long-range interaction refers to modeling dependencies between distant feature points or blocks within an image, which can significantly enhance the model's robustness. Both CNN and Transformer can establish long-range interactions through layering and patch calculations. However, the underlying mechanism of long-range interaction in visual space remains unclear. We propose the mode-locking theory as the underlying mechanism, which constrains the phase and wavelength relationship between waves to achieve mode-locked interference waveform. We verify this theory through simulation experiments and demonstrate the mode-locking pattern in real-world scene models. Our proposed theory of long-range interaction provides a comprehensive understanding of the mechanism behind this phenomenon in artificial neural networks. This theory can inspire the integration of the mode-locking pattern into models to enhance their robustness.
comment: 10 pages, 6 figures
☆ Semantic-Preserving Augmentation for Robust Image-Text Retrieval ICASSP 2023
Image text retrieval is a task to search for the proper textual descriptions of the visual world and vice versa. One challenge of this task is the vulnerability to input image and text corruptions. Such corruptions are often unobserved during the training, and degrade the retrieval model decision quality substantially. In this paper, we propose a novel image text retrieval technique, referred to as robust visual semantic embedding (RVSE), which consists of novel image-based and text-based augmentation techniques called semantic preserving augmentation for image (SPAugI) and text (SPAugT). Since SPAugI and SPAugT change the original data in a way that its semantic information is preserved, we enforce the feature extractors to generate semantic aware embedding vectors regardless of the corruption, improving the model robustness significantly. From extensive experiments using benchmark datasets, we show that RVSE outperforms conventional retrieval schemes in terms of image-text retrieval performance.
comment: Accepted to ICASSP 2023
☆ Human Pose Estimation from Ambiguous Pressure Recordings with Spatio-temporal Masked Transformers
Despite the impressive performance of vision-based pose estimators, they generally fail to perform well under adverse vision conditions and often don't satisfy the privacy demands of customers. As a result, researchers have begun to study tactile sensing systems as an alternative. However, these systems suffer from noisy and ambiguous recordings. To tackle this problem, we propose a novel solution for pose estimation from ambiguous pressure data. Our method comprises a spatio-temporal vision transformer with an encoder-decoder architecture. Detailed experiments on two popular public datasets reveal that our model outperforms existing solutions in the area. Moreover, we observe that increasing the number of temporal crops in the early stages of the network positively impacts the performance while pre-training the network in a self-supervised setting using a masked auto-encoder approach also further improves the results.
☆ Inducing Neural Collapse to a Fixed Hierarchy-Aware Frame for Reducing Mistake Severity
There is a recently discovered and intriguing phenomenon called Neural Collapse: at the terminal phase of training a deep neural network for classification, the within-class penultimate feature means and the associated classifier vectors of all flat classes collapse to the vertices of a simplex Equiangular Tight Frame (ETF). Recent work has tried to exploit this phenomenon by fixing the related classifier weights to a pre-computed ETF to induce neural collapse and maximize the separation of the learned features when training with imbalanced data. In this work, we propose to fix the linear classifier of a deep neural network to a Hierarchy-Aware Frame (HAFrame), instead of an ETF, and use a cosine similarity-based auxiliary loss to learn hierarchy-aware penultimate features that collapse to the HAFrame. We demonstrate that our approach reduces the mistake severity of the model's predictions while maintaining its top-1 accuracy on several datasets of varying scales with hierarchies of heights ranging from 3 to 12. We will release our code on GitHub in the near future.
☆ Generalized Diffusion MRI Denoising and Super-Resolution using Swin Transformers
Diffusion MRI is a non-invasive, in-vivo medical imaging method able to map tissue microstructure and structural connectivity of the human brain, as well as detect changes, such as brain development and injury, not visible by other clinical neuroimaging techniques. However, acquiring high signal-to-noise ratio (SNR) datasets with high angular and spatial sampling requires prohibitively long scan times, limiting usage in many important clinical settings, especially children, the elderly, and emergency patients with acute neurological disorders who might not be able to cooperate with the MRI scan without conscious sedation or general anesthesia. Here, we propose to use a Swin UNEt TRansformers (Swin UNETR) model, trained on augmented Human Connectome Project (HCP) data and conditioned on registered T1 scans, to perform generalized denoising and super-resolution of diffusion MRI invariant to acquisition parameters, patient populations, scanners, and sites. We qualitatively demonstrate super-resolution with artificially downsampled HCP data in normal adult volunteers. Our experiments on two other unrelated datasets, one of children with neurodevelopmental disorders and one of traumatic brain injury patients, show that our method demonstrates superior denoising despite wide data distribution shifts. Further improvement can be achieved via finetuning with just one additional subject. We apply our model to diffusion tensor (2nd order spherical harmonic) and higher-order spherical harmonic coefficient estimation and show results superior to current state-of-the-art methods. Our method can be used out-of-the-box or minimally finetuned to denoise and super-resolve a wide variety of diffusion MRI datasets. The code and model are publicly available at https://github.com/ucsfncl/dmri-swin.
☆ An Adaptive GViT for Gas Mixture Identification and Concentration Estimation
Estimating the composition and concentration of ambient gases is crucial for industrial gas safety. Even though other researchers have proposed some gas identification and con-centration estimation algorithms, these algorithms still suffer from severe flaws, particularly in fulfilling industry demands. One example is that the lengths of data collected in an industrial setting tend to vary. The conventional algorithm, yet, cannot be used to analyze the variant-length data effectively. Trimming the data will preserve only steady-state values, inevitably leading to the loss of vital information. The gas identification and concentration estimation model called GCN-ViT(GViT) is proposed in this paper; we view the sensor data to be a one-way chain that has only been downscaled to retain the majority of the original in-formation. The GViT model can directly utilize sensor ar-rays' variable-length real-time signal data as input. We validated the above model on a dataset of 12-hour uninterrupted monitoring of two randomly varying gas mixtures, CO-ethylene and methane-ethylene. The accuracy of gas identification can reach 97.61%, R2 of the pure gas concentration estimation is above 99.5% on average, and R2 of the mixed gas concentration estimation is above 95% on average.
☆ HumanBench: Towards General Human-centric Perception with Projector Assisted Pretraining CVPR2023
Human-centric perceptions include a variety of vision tasks, which have widespread industrial applications, including surveillance, autonomous driving, and the metaverse. It is desirable to have a general pretrain model for versatile human-centric downstream tasks. This paper forges ahead along this path from the aspects of both benchmark and pretraining methods. Specifically, we propose a \textbf{HumanBench} based on existing datasets to comprehensively evaluate on the common ground the generalization abilities of different pretraining methods on 19 datasets from 6 diverse downstream tasks, including person ReID, pose estimation, human parsing, pedestrian attribute recognition, pedestrian detection, and crowd counting. To learn both coarse-grained and fine-grained knowledge in human bodies, we further propose a \textbf{P}rojector \textbf{A}ssis\textbf{T}ed \textbf{H}ierarchical pretraining method (\textbf{PATH}) to learn diverse knowledge at different granularity levels. Comprehensive evaluations on HumanBench show that our PATH achieves new state-of-the-art results on 17 downstream datasets and on-par results on the other 2 datasets. The code will be publicly at \href{https://github.com/OpenGVLab/HumanBench}{https://github.com/OpenGVLab/HumanBench}.
comment: Accepted to CVPR2023
☆ Tag2Text: Guiding Vision-Language Model via Image Tagging
This paper presents Tag2Text, a vision language pre-training (VLP) framework, which introduces image tagging into vision-language models to guide the learning of visual-linguistic features. In contrast to prior works which utilize object tags either manually labeled or automatically detected with a limited detector, our approach utilizes tags parsed from its paired text to learn an image tagger and meanwhile provides guidance to vision-language models. Given that, Tag2Text can utilize large-scale annotation-free image tags in accordance with image-text pairs, and provides more diverse tag categories beyond objects. As a result, Tag2Text achieves a superior image tag recognition ability by exploiting fine-grained text information. Moreover, by leveraging tagging guidance, Tag2Text effectively enhances the performance of vision-language models on both generation-based and alignment-based tasks. Across a wide range of downstream benchmarks, Tag2Text achieves state-of-the-art or competitive results with similar model sizes and data scales, demonstrating the efficacy of the proposed tagging guidance.
☆ EHRDiff: Exploring Realistic EHR Synthesis with Diffusion Models
Electronic health records (EHR) contain vast biomedical knowledge and are rich resources for developing precise medicine systems. However, due to privacy concerns, there are limited high-quality EHR data accessible to researchers hence hindering the advancement of methodologies. Recent research has explored using generative modelling methods to synthesize realistic EHR data, and most proposed methods are based on the generative adversarial network (GAN) and its variants for EHR synthesis. Although GAN-style methods achieved state-of-the-art performance in generating high-quality EHR data, such methods are hard to train and prone to mode collapse. Diffusion models are recently proposed generative modelling methods and set cutting-edge performance in image generation. The performance of diffusion models in realistic EHR synthesis is rarely explored. In this work, we explore whether the superior performance of diffusion models can translate to the domain of EHR synthesis and propose a novel EHR synthesis method named EHRDiff. Through comprehensive experiments, EHRDiff achieves new state-of-the-art performance for the quality of synthetic EHR data and can better protect private information in real training EHRs in the meanwhile.
comment: Working in progress
☆ Direct Robot Configuration Space Construction using Convolutional Encoder-Decoders
Intelligent robots must be able to perform safe and efficient motion planning in their environments. Central to modern motion planning is the configuration space. Configuration spaces define the set of configurations of a robot that result in collisions with obstacles in the workspace, C-clsn, and the set of configurations that do not, C-free. Modern approaches to motion planning first compute the configuration space and then perform motion planning using the calculated configuration space. Real-time motion planning requires accurate and efficient construction of configuration spaces. We are the first to apply a convolutional encoder-decoder framework for calculating highly accurate approximations to configuration spaces. Our model achieves an average 97.5% F1-score for predicting C-free and C-clsn for 2-D robotic workspaces with a dual-arm robot. Our method limits undetected collisions to less than 2.5% on robotic workspaces that involve translation, rotation, and removal of obstacles. Our model learns highly transferable features between robotic workspaces, requiring little to no fine-tuning to adapt to new transformations of obstacles in the workspace.
comment: 6 pages, 7 figures, 4 tables
☆ GATOR: Graph-Aware Transformer with Motion-Disentangled Regression for Human Mesh Recovery from a 2D Pose ICASSP 2023
3D human mesh recovery from a 2D pose plays an important role in various applications. However, it is hard for existing methods to simultaneously capture the multiple relations during the evolution from skeleton to mesh, including joint-joint, joint-vertex and vertex-vertex relations, which often leads to implausible results. To address this issue, we propose a novel solution, called GATOR, that contains an encoder of Graph-Aware Transformer (GAT) and a decoder with Motion-Disentangled Regression (MDR) to explore these multiple relations. Specifically, GAT combines a GCN and a graph-aware self-attention in parallel to capture physical and hidden joint-joint relations. Furthermore, MDR models joint-vertex and vertex-vertex interactions to explore joint and vertex relations. Based on the clustering characteristics of vertex offset fields, MDR regresses the vertices by composing the predicted base motions. Extensive experiments show that GATOR achieves state-of-the-art performance on two challenging benchmarks.
comment: Accepted by ICASSP 2023
☆ Iterative Few-shot Semantic Segmentation from Image Label Text
Few-shot semantic segmentation aims to learn to segment unseen class objects with the guidance of only a few support images. Most previous methods rely on the pixel-level label of support images. In this paper, we focus on a more challenging setting, in which only the image-level labels are available. We propose a general framework to firstly generate coarse masks with the help of the powerful vision-language model CLIP, and then iteratively and mutually refine the mask predictions of support and query images. Extensive experiments on PASCAL-5i and COCO-20i datasets demonstrate that our method not only outperforms the state-of-the-art weakly supervised approaches by a significant margin, but also achieves comparable or better results to recent supervised methods. Moreover, our method owns an excellent generalization ability for the images in the wild and uncommon classes. Code will be available at https://github.com/Whileherham/IMR-HSNet.
comment: ijcai 2022
Self-Supervised One-Shot Learning for Automatic Segmentation of StyleGAN Images
We propose in this paper a framework for automatic one-shot segmentation of synthetic images generated using StyleGANs. As to the need for `one-shot segmentation', we want the network to carry out a semantic segmentation of the images on the fly, that is, as they are being produced at inference time. The implementation of our framework is based on the observation that the multi-scale hidden features produced by a GAN during image synthesis hold useful semantic information that can be utilized for automatic segmentation. Using these features, our proposed framework learns to segment synthetic images using a novel self-supervised, contrastive clustering algorithm that projects the hidden features in the generator onto a compact feature space for per-pixel classification. This contrastive learner uses a swapped prediction loss for image segmentation that is computed using pixel-wise cluster assignments for the image and its transformed variants. Using the hidden features from an already pre-trained GAN for clustering, this leads to a much faster learning of the pixel-wise feature vectors for one-shot segmentation. We have tested our implementation on a number of standard benchmarks (CelebA, LSUN, PASCAL-Part) for object and part segmentation. The results of our experiments yield a segmentation performance that not only outperforms the semi-supervised baseline methods with an average wIoU margin of 1.02 % but also improves the inference speeds by a peak factor of 4.5. Finally, we also show the results of using the proposed framework in the implementation of BagGAN, a GAN-based framework for the production of annotated synthetic baggage X-ray scans for threat detection. This one-shot learning framework was trained and tested on the PIDRay baggage screening benchmark for 5 different threat categories to yield a segmentation performance which stands close to its baseline segmenter.
☆ Fusarium head blight detection, spikelet estimation, and severity assessment in wheat using 3D convolutional neural networks
Fusarium head blight (FHB) is one of the most significant diseases affecting wheat and other small grain cereals worldwide. The development of resistant varieties requires the laborious task of field and greenhouse phenotyping. The applications considered in this work are the automated detection of FHB disease symptoms expressed on a wheat plant, the automated estimation of the total number of spikelets and the total number of infected spikelets on a wheat head, and the automated assessment of the FHB severity in infected wheat. The data used to generate the results are 3-dimensional (3D) multispectral point clouds (PC), which are 3D collections of points - each associated with a red, green, blue (RGB), and near-infrared (NIR) measurement. Over 300 wheat plant images were collected using a multispectral 3D scanner, and the labelled UW-MRDC 3D wheat dataset was created. The data was used to develop novel and efficient 3D convolutional neural network (CNN) models for FHB detection, which achieved 100% accuracy. The influence of the multispectral information on performance was evaluated, and our results showed the dominance of the RGB channels over both the NIR and the NIR plus RGB channels combined. Furthermore, novel and efficient 3D CNNs were created to estimate the total number of spikelets and the total number of infected spikelets on a wheat head, and our best models achieved mean absolute errors (MAE) of 1.13 and 1.56, respectively. Moreover, 3D CNN models for FHB severity estimation were created, and our best model achieved 8.6 MAE. A linear regression analysis between the visual FHB severity assessment and the FHB severity predicted by our 3D CNN was performed, and the results showed a significant correlation between the two variables with a 0.0001 P-value and 0.94 R-squared.
♻ ☆ DDPNAS: Efficient Neural Architecture Search via Dynamic Distribution Pruning
Neural Architecture Search (NAS) has demonstrated state-of-the-art performance on various computer vision tasks. Despite the superior performance achieved, the efficiency and generality of existing methods are highly valued due to their high computational complexity and low generality. In this paper, we propose an efficient and unified NAS framework termed DDPNAS via dynamic distribution pruning, facilitating a theoretical bound on accuracy and efficiency. In particular, we first sample architectures from a joint categorical distribution. Then the search space is dynamically pruned and its distribution is updated every few epochs. With the proposed efficient network generation method, we directly obtain the optimal neural architectures on given constraints, which is practical for on-device models across diverse search spaces and constraints. The architectures searched by our method achieve remarkable top-1 accuracies, 97.56 and 77.2 on CIFAR-10 and ImageNet (mobile settings), respectively, with the fastest search process, i.e., only 1.8 GPU hours on a Tesla V100. Codes for searching and network generation are available at: https://openi.pcl.ac.cn/PCL AutoML/XNAS.
comment: A update version of this work. 19 pages
♻ ☆ DEJA VU: Continual Model Generalization For Unseen Domains ICLR 2023
In real-world applications, deep learning models often run in non-stationary environments where the target data distribution continually shifts over time. There have been numerous domain adaptation (DA) methods in both online and offline modes to improve cross-domain adaptation ability. However, these DA methods typically only provide good performance after a long period of adaptation, and perform poorly on new domains before and during adaptation - in what we call the "Unfamiliar Period", especially when domain shifts happen suddenly and significantly. On the other hand, domain generalization (DG) methods have been proposed to improve the model generalization ability on unadapted domains. However, existing DG works are ineffective for continually changing domains due to severe catastrophic forgetting of learned knowledge. To overcome these limitations of DA and DG in handling the Unfamiliar Period during continual domain shift, we propose RaTP, a framework that focuses on improving models' target domain generalization (TDG) capability, while also achieving effective target domain adaptation (TDA) capability right after training on certain domains and forgetting alleviation (FA) capability on past domains. RaTP includes a training-free data augmentation module to prepare data for TDG, a novel pseudo-labeling mechanism to provide reliable supervision for TDA, and a prototype contrastive alignment algorithm to align different domains for achieving TDG, TDA and FA. Extensive experiments on Digits, PACS, and DomainNet demonstrate that RaTP significantly outperforms state-of-the-art works from Continual DA, Source-Free DA, Test-Time/Online DA, Single DG, Multiple DG and Unified DA&DG in TDG, and achieves comparable TDA and FA capabilities.
comment: Published as a conference paper at ICLR 2023
♻ ☆ Nerflets: Local Radiance Fields for Efficient Structure-Aware 3D Scene Representation from 2D Supervision CVPR 2023
We address efficient and structure-aware 3D scene representation from images. Nerflets are our key contribution -- a set of local neural radiance fields that together represent a scene. Each nerflet maintains its own spatial position, orientation, and extent, within which it contributes to panoptic, density, and radiance reconstructions. By leveraging only photometric and inferred panoptic image supervision, we can directly and jointly optimize the parameters of a set of nerflets so as to form a decomposed representation of the scene, where each object instance is represented by a group of nerflets. During experiments with indoor and outdoor environments, we find that nerflets: (1) fit and approximate the scene more efficiently than traditional global NeRFs, (2) allow the extraction of panoptic and photometric renderings from arbitrary views, and (3) enable tasks rare for NeRFs, such as 3D panoptic segmentation and interactive editing.
comment: accepted by CVPR 2023
♻ ☆ APTx: better activation function than MISH, SWISH, and ReLU's variants used in deep learning
Activation Functions introduce non-linearity in the deep neural networks. This nonlinearity helps the neural networks learn faster and efficiently from the dataset. In deep learning, many activation functions are developed and used based on the type of problem statement. ReLU's variants, SWISH, and MISH are goto activation functions. MISH function is considered having similar or even better performance than SWISH, and much better than ReLU. In this paper, we propose an activation function named APTx which behaves similar to MISH, but requires lesser mathematical operations to compute. The lesser computational requirements of APTx does speed up the model training, and thus also reduces the hardware requirement for the deep learning model.
comment: 8 pages, 6 figures
POLICE: Provably Optimal Linear Constraint Enforcement for Deep Neural Networks
Deep Neural Networks (DNNs) outshine alternative function approximators in many settings thanks to their modularity in composing any desired differentiable operator. The formed parametrized functional is then tuned to solve a task at hand from simple gradient descent. This modularity comes at the cost of making strict enforcement of constraints on DNNs, e.g. from a priori knowledge of the task, or from desired physical properties, an open challenge. In this paper we propose the first provable affine constraint enforcement method for DNNs that only requires minimal changes into a given DNN's forward-pass, that is computationally friendly, and that leaves the optimization of the DNN's parameter to be unconstrained, i.e. standard gradient-based method can be employed. Our method does not require any sampling and provably ensures that the DNN fulfills the affine constraint on a given input space's region at any point during training, and testing. We coin this method POLICE, standing for Provably Optimal LInear Constraint Enforcement. Github: https://github.com/RandallBalestriero/POLICE
♻ ☆ Human-machine Interactive Tissue Prototype Learning for Label-efficient Histopathology Image Segmentation
Recently, deep neural networks have greatly advanced histopathology image segmentation but usually require abundant annotated data. However, due to the gigapixel scale of whole slide images and pathologists' heavy daily workload, obtaining pixel-level labels for supervised learning in clinical practice is often infeasible. Alternatively, weakly-supervised segmentation methods have been explored with less laborious image-level labels, but their performance is unsatisfactory due to the lack of dense supervision. Inspired by the recent success of self-supervised learning methods, we present a label-efficient tissue prototype dictionary building pipeline and propose to use the obtained prototypes to guide histopathology image segmentation. Particularly, taking advantage of self-supervised contrastive learning, an encoder is trained to project the unlabeled histopathology image patches into a discriminative embedding space where these patches are clustered to identify the tissue prototypes by efficient pathologists' visual examination. Then, the encoder is used to map the images into the embedding space and generate pixel-level pseudo tissue masks by querying the tissue prototype dictionary. Finally, the pseudo masks are used to train a segmentation network with dense supervision for better performance. Experiments on two public datasets demonstrate that our human-machine interactive tissue prototype learning method can achieve comparable segmentation performance as the fully-supervised baselines with less annotation burden and outperform other weakly-supervised methods. Codes will be available upon publication.
comment: IPMI2023 camera ready
♻ ☆ DORA: Exploring outlier representations in Deep Neural Networks
Deep Neural Networks (DNNs) draw their power from the representations they learn. However, while being incredibly effective in learning complex abstractions, they are susceptible to learning malicious concepts, due to the spurious correlations inherent in the training data. So far, existing methods for uncovering such artifactual behavior in trained models focus on finding artifacts in the input data, which requires both availability of a data set and human supervision. In this paper, we introduce DORA (Data-agnOstic Representation Analysis): the first data-agnostic framework for the analysis of the representation space of DNNs. We propose a novel distance measure between representations that utilizes self-explaining capabilities within the network itself without access to any data and quantitatively validate its alignment with human-defined semantic distances. We further demonstrate that this metric could be utilized for the detection of anomalous representations, which may bear a risk of learning unintended spurious concepts deviating from the desired decision-making policy. Finally, we demonstrate the practical utility of DORA by analyzing and identifying artifactual representations in widely popular Computer Vision models.
comment: 15 pages, 11 figures
♻ ☆ StructVPR: Distill Structural Knowledge with Weighting Samples for Visual Place Recognition
Visual place recognition (VPR) is usually considered as a specific image retrieval problem. Limited by existing training frameworks, most deep learning-based works cannot extract sufficiently stable global features from RGB images and rely on a time-consuming re-ranking step to exploit spatial structural information for better performance. In this paper, we propose StructVPR, a novel training architecture for VPR, to enhance structural knowledge in RGB global features and thus improve feature stability in a constantly changing environment. Specifically, StructVPR uses segmentation images as a more definitive source of structural knowledge input into a CNN network and applies knowledge distillation to avoid online segmentation and inference of seg-branch in testing. Considering that not all samples contain high-quality and helpful knowledge, and some even hurt the performance of distillation, we partition samples and weigh each sample's distillation loss to enhance the expected knowledge precisely. Finally, StructVPR achieves impressive performance on several benchmarks using only global retrieval and even outperforms many two-stage approaches by a large margin. After adding additional re-ranking, ours achieves state-of-the-art performance while maintaining a low computational cost.
♻ ☆ GradMA: A Gradient-Memory-based Accelerated Federated Learning with Alleviated Catastrophic Forgetting
Federated Learning (FL) has emerged as a de facto machine learning area and received rapid increasing research interests from the community. However, catastrophic forgetting caused by data heterogeneity and partial participation poses distinctive challenges for FL, which are detrimental to the performance. To tackle the problems, we propose a new FL approach (namely GradMA), which takes inspiration from continual learning to simultaneously correct the server-side and worker-side update directions as well as take full advantage of server's rich computing and memory resources. Furthermore, we elaborate a memory reduction strategy to enable GradMA to accommodate FL with a large scale of workers. We then analyze convergence of GradMA theoretically under the smooth non-convex setting and show that its convergence rate achieves a linear speed up w.r.t the increasing number of sampled active workers. At last, our extensive experiments on various image classification tasks show that GradMA achieves significant performance gains in accuracy and communication efficiency compared to SOTA baselines.
♻ ☆ Occlusion-Aware Instance Segmentation via BiLayer Network Architectures CVPR 2021
Segmenting highly-overlapping image objects is challenging, because there is typically no distinction between real object contours and occlusion boundaries on images. Unlike previous instance segmentation methods, we model image formation as a composition of two overlapping layers, and propose Bilayer Convolutional Network (BCNet), where the top layer detects occluding objects (occluders) and the bottom layer infers partially occluded instances (occludees). The explicit modeling of occlusion relationship with bilayer structure naturally decouples the boundaries of both the occluding and occluded instances, and considers the interaction between them during mask regression. We investigate the efficacy of bilayer structure using two popular convolutional network designs, namely, Fully Convolutional Network (FCN) and Graph Convolutional Network (GCN). Further, we formulate bilayer decoupling using the vision transformer (ViT), by representing instances in the image as separate learnable occluder and occludee queries. Large and consistent improvements using one/two-stage and query-based object detectors with various backbones and network layer choices validate the generalization ability of bilayer decoupling, as shown by extensive experiments on image instance segmentation benchmarks (COCO, KINS, COCOA) and video instance segmentation benchmarks (YTVIS, OVIS, BDD100K MOTS), especially for heavy occlusion cases. Code and data are available at https://github.com/lkeab/BCNet.
comment: Extended version of "Deep Occlusion-Aware Instance Segmentation with Overlapping BiLayers", CVPR 2021 (arXiv:2103.12340)
♻ ☆ PPCR: Learning Pyramid Pixel Context Recalibration Module for Medical Image Classification
Spatial attention mechanism has been widely incorporated into deep convolutional neural networks (CNNs) via long-range dependency capturing, significantly lifting the performance in computer vision, but it may perform poorly in medical imaging. Unfortunately, existing efforts are often unaware that long-range dependency capturing has limitations in highlighting subtle lesion regions, neglecting to exploit the potential of multi-scale pixel context information to improve the representational capability of CNNs. In this paper, we propose a practical yet lightweight architectural unit, Pyramid Pixel Context Recalibration (PPCR) module, which exploits multi-scale pixel context information to recalibrate pixel position in a pixel-independent manner adaptively. PPCR first designs a cross-channel pyramid pooling to aggregate multi-scale pixel context information, then eliminates the inconsistency among them by the well-designed pixel normalization, and finally estimates per pixel attention weight via a pixel context integration. PPCR can be flexibly plugged into modern CNNs with negligible overhead. Extensive experiments on five medical image datasets and CIFAR benchmarks empirically demonstrate the superiority and generalization of PPCR over state-of-the-art attention methods. The in-depth analyses explain the inherent behavior of PPCR in the decision-making process, improving the interpretability of CNNs.
comment: 10 pages
♻ ☆ Exploiting Proximity-Aware Tasks for Embodied Social Navigation
Learning how to navigate among humans in an occluded and spatially constrained indoor environment, is a key ability required to embodied agent to be integrated into our society. In this paper, we propose an end-to-end architecture that exploits Proximity-Aware Tasks (referred as to Risk and Proximity Compass) to inject into a reinforcement learning navigation policy the ability to infer common-sense social behaviors. To this end, our tasks exploit the notion of immediate and future dangers of collision. Furthermore, we propose an evaluation protocol specifically designed for the Social Navigation Task in simulated environments. This is done to capture fine-grained features and characteristics of the policy by analyzing the minimal unit of human-robot spatial interaction, called Encounter. We validate our approach on Gibson4+ and Habitat-Matterport3D datasets.
♻ ☆ EiX-GNN : Concept-level eigencentrality explainer for graph neural networks
Nowadays, deep prediction models, especially graph neural networks, have a majorplace in critical applications. In such context, those models need to be highlyinterpretable or being explainable by humans, and at the societal scope, this understandingmay also be feasible for humans that do not have a strong prior knowledgein models and contexts that need to be explained. In the literature, explainingis a human knowledge transfer process regarding a phenomenon between an explainerand an explainee. We propose EiX-GNN (Eigencentrality eXplainer forGraph Neural Networks) a new powerful method for explaining graph neural networksthat encodes computationally this social explainer-to-explainee dependenceunderlying in the explanation process. To handle this dependency, we introducethe notion of explainee concept assimibility which allows explainer to adapt itsexplanation to explainee background or expectation. We lead a qualitative studyto illustrate our explainee concept assimibility notion on real-world data as wellas a qualitative study that compares, according to objective metrics established inthe literature, fairness and compactness of our method with respect to performingstate-of-the-art methods. It turns out that our method achieves strong results inboth aspects.
♻ ☆ LocPoseNet: Robust Location Prior for Unseen Object Pose Estimation
Object location priors have been shown to be critical for the standard 6D object pose estimation setting, where the training and testing objects are the same. Specifically, they can be used to initialize the 3D object translation and facilitate 3D object rotation estimation. Unfortunately, the object detectors that are used for this purpose do not generalize to unseen objects, i.e., objects from new categories at test time. Therefore, existing 6D pose estimation methods for previously-unseen objects either assume the ground-truth object location to be known, or yield inaccurate results when it is unavailable. In this paper, we address this problem by developing a method, LocPoseNet, able to robustly learn location prior for unseen objects. Our method builds upon a template matching strategy, where we propose to distribute the reference kernels and convolve them with a query to efficiently compute multi-scale correlations. We then introduce a novel translation estimator, which decouples scale-aware and scale-robust features to predict different object location parameters. Our method outperforms existing works by a large margin on LINEMOD and GenMOP. We further construct a challenging synthetic dataset, which allows us to highlight the better robustness of our method to various noise sources.
♻ ☆ No Reason for No Supervision: Improved Generalization in Supervised Models ICLR 2023
We consider the problem of training a deep neural network on a given classification task, e.g., ImageNet-1K (IN1K), so that it excels at both the training task as well as at other (future) transfer tasks. These two seemingly contradictory properties impose a trade-off between improving the model's generalization and maintaining its performance on the original task. Models trained with self-supervised learning tend to generalize better than their supervised counterparts for transfer learning; yet, they still lag behind supervised models on IN1K. In this paper, we propose a supervised learning setup that leverages the best of both worlds. We extensively analyze supervised training using multi-scale crops for data augmentation and an expendable projector head, and reveal that the design of the projector allows us to control the trade-off between performance on the training task and transferability. We further replace the last layer of class weights with class prototypes computed on the fly using a memory bank and derive two models: t-ReX that achieves a new state of the art for transfer learning and outperforms top methods such as DINO and PAWS on IN1K, and t-ReX* that matches the highly optimized RSB-A1 model on IN1K while performing better on transfer tasks. Code and pretrained models: https://europe.naverlabs.com/t-rex
comment: Accepted to ICLR 2023 (spotlight)
♻ ☆ Weakly Supervised Object Localization via Transformer with Implicit Spatial Calibration ECCV2022
Weakly Supervised Object Localization (WSOL), which aims to localize objects by only using image-level labels, has attracted much attention because of its low annotation cost in real applications. Recent studies leverage the advantage of self-attention in visual Transformer for long-range dependency to re-active semantic regions, aiming to avoid partial activation in traditional class activation mapping (CAM). However, the long-range modeling in Transformer neglects the inherent spatial coherence of the object, and it usually diffuses the semantic-aware regions far from the object boundary, making localization results significantly larger or far smaller. To address such an issue, we introduce a simple yet effective Spatial Calibration Module (SCM) for accurate WSOL, incorporating semantic similarities of patch tokens and their spatial relationships into a unified diffusion model. Specifically, we introduce a learnable parameter to dynamically adjust the semantic correlations and spatial context intensities for effective information propagation. In practice, SCM is designed as an external module of Transformer, and can be removed during inference to reduce the computation cost. The object-sensitive localization ability is implicitly embedded into the Transformer encoder through optimization in the training phase. It enables the generated attention maps to capture the sharper object boundaries and filter the object-irrelevant background area. Extensive experimental results demonstrate the effectiveness of the proposed method, which significantly outperforms its counterpart TS-CAM on both CUB-200 and ImageNet-1K benchmarks. The code is available at https://github.com/164140757/SCM.
comment: Accepted by ECCV2022
♻ ☆ GroupGazer: A Tool to Compute the Gaze per Participant in Groups with integrated Calibration to Map the Gaze Online to a Screen or Beamer Projection
In this paper we present GroupGaze. It is a tool that can be used to calculate the gaze direction and the gaze position of whole groups. GroupGazer calculates the gaze direction of every single person in the image and allows to map these gaze vectors to a projection like a projector. In addition to the person-specific gaze direction, the person affiliation of each gaze vector is stored based on the position in the image. Also, it is possible to save the group attention after a calibration. The software is free to use and requires a simple webcam as well as an NVIDIA GPU and the operating system Windows or Linux. Link: https://es-cloud.cs.uni-tuebingen.de/d/8e2ab8c3fdd444e1a135/?p=%2FGroupGazer&mode=list
♻ ☆ Robotic Fabric Flattening with Wrinkle Direction Detection
Deformable Object Manipulation (DOM) is an important field of research as it contributes to practical tasks such as automatic cloth handling, cable routing, surgical operation, etc. Perception is considered one of the major challenges in DOM due to the complex dynamics and high degree of freedom of deformable objects. In this paper, we develop a novel image-processing algorithm based on Gabor filters to extract useful features from cloth, and based on this, devise a strategy for cloth flattening tasks. We evaluate the overall framework experimentally, and compare it with three human operators. The results show that our algorithm can determine the direction of wrinkles on the cloth accurately in the simulation as well as the real robot experiments. Besides, the robot executing the flattening tasks using the dewrinkling strategy given by our algorithm achieves satisfying performance compared to other baseline methods. The experiment video is available on https://sites.google.com/view/robotic-fabric-flattening/home
♻ ☆ Pistol: Pupil Invisible Supportive Tool to extract Pupil, Iris, Eye Opening, Eye Movements, Pupil and Iris Gaze Vector, and 2D as well as 3D Gaze
This paper describes a feature extraction and gaze estimation software, named \textit{Pistol} that can be used with Pupil Invisible projects and other eye trackers in the future. In offline mode, our software extracts multiple features from the eye including, the pupil and iris ellipse, eye aperture, pupil vector, iris vector, eye movement types from pupil and iris velocities, marker detection, marker distance, 2D gaze estimation for the pupil center, iris center, pupil vector, and iris vector using Levenberg Marquart fitting and neural networks. The gaze signal is computed in 2D for each eye and each feature separately and for both eyes in 3D also for each feature separately. We hope this software helps other researchers to extract state-of-the-art features for their research out of their recordings. Link: https://es-cloud.cs.uni-tuebingen.de/d/8e2ab8c3fdd444e1a135/?p=%2FPISTOL&mode=list
♻ ☆ The Gaze and Mouse Signal as additional Source for User Fingerprints in Browser Applications
In this work, we inspect different data sources for browser fingerprints. We show which disadvantages and limitations browser statistics have and how this can be avoided with other data sources. Since human visual behavior is a rich source of information and also contains person specific information, it is a valuable source for browser fingerprints. However, human gaze acquisition in the browser also has disadvantages, such as inaccuracies via webcam and the restriction that the user must first allow access to the camera. However, it is also known that the mouse movements and the human gaze correlate and therefore, the mouse movements can be used instead of the gaze signal. In our evaluation, we show the influence of all possible combinations of the three information sources for user recognition and describe our simple approach in detail. Link: https://es-cloud.cs.uni-tuebingen.de/d/8e2ab8c3fdd444e1a135/?p=%2FThe%20Gaze%20and%20Mouse%20Signal%20as%20additional%20Source%20...&mode=list
♻ ☆ You Only Need End-to-End Training for Long-Tailed Recognition
The generalization gap on the long-tailed data sets is largely owing to most categories only occupying a few training samples. Decoupled training achieves better performance by training backbone and classifier separately. What causes the poorer performance of end-to-end model training (e.g., logits margin-based methods)? In this work, we identify a key factor that affects the learning of the classifier: the channel-correlated features with low entropy before inputting into the classifier. From the perspective of information theory, we analyze why cross-entropy loss tends to produce highly correlated features on the imbalanced data. In addition, we theoretically analyze and prove its impacts on the gradients of classifier weights, the condition number of Hessian, and logits margin-based approach. Therefore, we firstly propose to use Channel Whitening to decorrelate ("scatter") the classifier's inputs for decoupling the weight update and reshaping the skewed decision boundary, which achieves satisfactory results combined with logits margin-based method. However, when the number of minor classes are large, batch imbalance and more participation in training cause over-fitting of the major classes. We also propose two novel modules, Block-based Relatively Balanced Batch Sampler (B3RS) and Batch Embedded Training (BET) to solve the above problems, which makes the end-to-end training achieve even better performance than decoupled training. Experimental results on the long-tailed classification benchmarks, CIFAR-LT and ImageNet-LT, demonstrate the effectiveness of our method.
comment: This is a draft
♻ ☆ Data-Efficient Training of CNNs and Transformers with Coresets: A Stability Perspective
Coreset selection is among the most effective ways to reduce the training time of CNNs, however, only limited is known on how the resultant models will behave under variations of the coreset size, and choice of datasets and models. Moreover, given the recent paradigm shift towards transformer-based models, it is still an open question how coreset selection would impact their performance. There are several similar intriguing questions that need to be answered for a wide acceptance of coreset selection methods, and this paper attempts to answer some of these. We present a systematic benchmarking setup and perform a rigorous comparison of different coreset selection methods on CNNs and transformers. Our investigation reveals that under certain circumstances, random selection of subsets is more robust and stable when compared with the SOTA selection methods. We demonstrate that the conventional concept of uniform subset sampling across the various classes of the data is not the appropriate choice. Rather samples should be adaptively chosen based on the complexity of the data distribution for each class. Transformers are generally pretrained on large datasets, and we show that for certain target datasets, it helps to keep their performance stable at even very small coreset sizes. We further show that when no pretraining is done or when the pretrained transformer models are used with non-natural images (e.g. medical data), CNNs tend to generalize better than transformers at even very small coreset sizes. Lastly, we demonstrate that in the absence of the right pretraining, CNNs are better at learning the semantic coherence between spatially distant objects within an image, and these tend to outperform transformers at almost all choices of the coreset size.
♻ ☆ Neural Video Compression with Diverse Contexts CVPR 2023
For any video codecs, the coding efficiency highly relies on whether the current signal to be encoded can find the relevant contexts from the previous reconstructed signals. Traditional codec has verified more contexts bring substantial coding gain, but in a time-consuming manner. However, for the emerging neural video codec (NVC), its contexts are still limited, leading to low compression ratio. To boost NVC, this paper proposes increasing the context diversity in both temporal and spatial dimensions. First, we guide the model to learn hierarchical quality patterns across frames, which enriches long-term and yet high-quality temporal contexts. Furthermore, to tap the potential of optical flow-based coding framework, we introduce a group-based offset diversity where the cross-group interaction is proposed for better context mining. In addition, this paper also adopts a quadtree-based partition to increase spatial context diversity when encoding the latent representation in parallel. Experiments show that our codec obtains 23.5% bitrate saving over previous SOTA NVC. Better yet, our codec has surpassed the under-developing next generation traditional codec/ECM in both RGB and YUV420 colorspaces, in terms of PSNR. The codes are at https://github.com/microsoft/DCVC.
comment: Accepted by CVPR 2023. Codes are at https://github.com/microsoft/DCVC
♻ ☆ BaDLAD: A Large Multi-Domain Bengali Document Layout Analysis Dataset
While strides have been made in deep learning based Bengali Optical Character Recognition (OCR) in the past decade, the absence of large Document Layout Analysis (DLA) datasets has hindered the application of OCR in document transcription, e.g., transcribing historical documents and newspapers. Moreover, rule-based DLA systems that are currently being employed in practice are not robust to domain variations and out-of-distribution layouts. To this end, we present the first multidomain large Bengali Document Layout Analysis Dataset: BaDLAD. This dataset contains 33,695 human annotated document samples from six domains - i) books and magazines, ii) public domain govt. documents, iii) liberation war documents, iv) newspapers, v) historical newspapers, and vi) property deeds, with 710K polygon annotations for four unit types: text-box, paragraph, image, and table. Through preliminary experiments benchmarking the performance of existing state-of-the-art deep learning architectures for English DLA, we demonstrate the efficacy of our dataset in training deep learning based Bengali document digitization models.
♻ ☆ Contrastive Audio-Visual Masked Autoencoder ICLR 2023
In this paper, we first extend the recent Masked Auto-Encoder (MAE) model from a single modality to audio-visual multi-modalities. Subsequently, we propose the Contrastive Audio-Visual Masked Auto-Encoder (CAV-MAE) by combining contrastive learning and masked data modeling, two major self-supervised learning frameworks, to learn a joint and coordinated audio-visual representation. Our experiments show that the contrastive audio-visual correspondence learning objective not only enables the model to perform audio-visual retrieval tasks, but also helps the model learn a better joint representation. As a result, our fully self-supervised pretrained CAV-MAE achieves a new SOTA accuracy of 65.9% on VGGSound, and is comparable with the previous best supervised pretrained model on AudioSet in the audio-visual event classification task. Code and pretrained models are at https://github.com/yuangongnd/cav-mae.
comment: Accepted at ICLR 2023 as a notable top 25% paper. Code and pretrained models are at https://github.com/yuangongnd/cav-mae
♻ ☆ A survey on automated detection and classification of acute leukemia and WBCs in microscopic blood cells
Leukemia (blood cancer) is an unusual spread of White Blood Cells or Leukocytes (WBCs) in the bone marrow and blood. Pathologists can diagnose leukemia by looking at a person's blood sample under a microscope. They identify and categorize leukemia by counting various blood cells and morphological features. This technique is time-consuming for the prediction of leukemia. The pathologist's professional skills and experiences may be affecting this procedure, too. In computer vision, traditional machine learning and deep learning techniques are practical roadmaps that increase the accuracy and speed in diagnosing and classifying medical images such as microscopic blood cells. This paper provides a comprehensive analysis of the detection and classification of acute leukemia and WBCs in the microscopic blood cells. First, we have divided the previous works into six categories based on the output of the models. Then, we describe various steps of detection and classification of acute leukemia and WBCs, including Data Augmentation, Preprocessing, Segmentation, Feature Extraction, Feature Selection (Reduction), Classification, and focus on classification step in the methods. Finally, we divide automated detection and classification of acute leukemia and WBCs into three categories, including traditional, Deep Neural Network (DNN), and mixture (traditional and DNN) methods based on the type of classifier in the classification step and analyze them. The results of this study show that in the diagnosis and classification of acute leukemia and WBCs, the Support Vector Machine (SVM) classifier in traditional machine learning models and Convolutional Neural Network (CNN) classifier in deep learning models have widely employed. The performance metrics of the models that use these classifiers compared to the others model are higher.
♻ ☆ DM-NeRF: 3D Scene Geometry Decomposition and Manipulation from 2D Images ICLR 2023
In this paper, we study the problem of 3D scene geometry decomposition and manipulation from 2D views. By leveraging the recent implicit neural representation techniques, particularly the appealing neural radiance fields, we introduce an object field component to learn unique codes for all individual objects in 3D space only from 2D supervision. The key to this component is a series of carefully designed loss functions to enable every 3D point, especially in non-occupied space, to be effectively optimized even without 3D labels. In addition, we introduce an inverse query algorithm to freely manipulate any specified 3D object shape in the learned scene representation. Notably, our manipulation algorithm can explicitly tackle key issues such as object collisions and visual occlusions. Our method, called DM-NeRF, is among the first to simultaneously reconstruct, decompose, manipulate and render complex 3D scenes in a single pipeline. Extensive experiments on three datasets clearly show that our method can accurately decompose all 3D objects from 2D views, allowing any interested object to be freely manipulated in 3D space such as translation, rotation, size adjustment, and deformation.
comment: ICLR 2023. Our data and code are available at: https://github.com/vLAR-group/DM-NeRF
♻ ☆ Neuro-Modulated Hebbian Learning for Fully Test-Time Adaptation CVPR2023
Fully test-time adaptation aims to adapt the network model based on sequential analysis of input samples during the inference stage to address the cross-domain performance degradation problem of deep neural networks. We take inspiration from the biological plausibility learning where the neuron responses are tuned based on a local synapse-change procedure and activated by competitive lateral inhibition rules. Based on these feed-forward learning rules, we design a soft Hebbian learning process which provides an unsupervised and effective mechanism for online adaptation. We observe that the performance of this feed-forward Hebbian learning for fully test-time adaptation can be significantly improved by incorporating a feedback neuro-modulation layer. It is able to fine-tune the neuron responses based on the external feedback generated by the error back-propagation from the top inference layers. This leads to our proposed neuro-modulated Hebbian learning (NHL) method for fully test-time adaptation. With the unsupervised feed-forward soft Hebbian learning being combined with a learned neuro-modulator to capture feedback from external responses, the source model can be effectively adapted during the testing process. Experimental results on benchmark datasets demonstrate that our proposed method can significantly improve the adaptation performance of network models and outperforms existing state-of-the-art methods.
comment: CVPR2023 accepted
♻ ☆ Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detection
In this paper, we present an open-set object detector, called Grounding DINO, by marrying Transformer-based detector DINO with grounded pre-training, which can detect arbitrary objects with human inputs such as category names or referring expressions. The key solution of open-set object detection is introducing language to a closed-set detector for open-set concept generalization. To effectively fuse language and vision modalities, we conceptually divide a closed-set detector into three phases and propose a tight fusion solution, which includes a feature enhancer, a language-guided query selection, and a cross-modality decoder for cross-modality fusion. While previous works mainly evaluate open-set object detection on novel categories, we propose to also perform evaluations on referring expression comprehension for objects specified with attributes. Grounding DINO performs remarkably well on all three settings, including benchmarks on COCO, LVIS, ODinW, and RefCOCO/+/g. Grounding DINO achieves a $52.5$ AP on the COCO detection zero-shot transfer benchmark, i.e., without any training data from COCO. It sets a new record on the ODinW zero-shot benchmark with a mean $26.1$ AP. Code will be available at \url{https://github.com/IDEA-Research/GroundingDINO}.
comment: Code will be available at https://github.com/IDEA-Research/GroundingDINO
♻ ☆ Parallel Inversion of Neural Radiance Fields for Robust Pose Estimation ICRA 2023
We present a parallelized optimization method based on fast Neural Radiance Fields (NeRF) for estimating 6-DoF pose of a camera with respect to an object or scene. Given a single observed RGB image of the target, we can predict the translation and rotation of the camera by minimizing the residual between pixels rendered from a fast NeRF model and pixels in the observed image. We integrate a momentum-based camera extrinsic optimization procedure into Instant Neural Graphics Primitives, a recent exceptionally fast NeRF implementation. By introducing parallel Monte Carlo sampling into the pose estimation task, our method overcomes local minima and improves efficiency in a more extensive search space. We also show the importance of adopting a more robust pixel-based loss function to reduce error. Experiments demonstrate that our method can achieve improved generalization and robustness on both synthetic and real-world benchmarks.
comment: ICRA 2023. Project page at https://pnerfp.github.io/
♻ ☆ Exploring Discontinuity for Video Frame Interpolation CVPR2023
Video frame interpolation (VFI) is the task that synthesizes the intermediate frame given two consecutive frames. Most of the previous studies have focused on appropriate frame warping operations and refinement modules for the warped frames. These studies have been conducted on natural videos containing only continuous motions. However, many practical videos contain various unnatural objects with discontinuous motions such as logos, user interfaces and subtitles. We propose three techniques to make the existing deep learning-based VFI architectures robust to these elements. First is a novel data augmentation strategy called figure-text mixing (FTM) which can make the models learn discontinuous motions during training stage without any extra dataset. Second, we propose a simple but effective module that predicts a map called discontinuity map (D-map), which densely distinguishes between areas of continuous and discontinuous motions. Lastly, we propose loss functions to give supervisions of the discontinuous motion areas which can be applied along with FTM and D-map. We additionally collect a special test benchmark called Graphical Discontinuous Motion (GDM) dataset consisting of some mobile games and chatting videos. Applied to the various state-of-the-art VFI networks, our method significantly improves the interpolation qualities on the videos from not only GDM dataset, but also the existing benchmarks containing only continuous motions such as Vimeo90K, UCF101, and DAVIS.
comment: CVPR2023 Accepted (The IEEE/CVF Conference on Computer Vision and Pattern Recognition 2023)
♻ ☆ RGB-Only Reconstruction of Tabletop Scenes for Collision-Free Manipulator Control ICRA 2023
We present a system for collision-free control of a robot manipulator that uses only RGB views of the world. Perceptual input of a tabletop scene is provided by multiple images of an RGB camera (without depth) that is either handheld or mounted on the robot end effector. A NeRF-like process is used to reconstruct the 3D geometry of the scene, from which the Euclidean full signed distance function (ESDF) is computed. A model predictive control algorithm is then used to control the manipulator to reach a desired pose while avoiding obstacles in the ESDF. We show results on a real dataset collected and annotated in our lab.
comment: ICRA 2023. Project page at https://ngp-mpc.github.io/
♻ ☆ A numerical approximation method for the Fisher-Rao distance between multivariate normal distributions
We present a simple method to approximate Rao's distance between multivariate normal distributions based on discretizing curves joining normal distributions and approximating Rao distances between successive nearby normal distributions on the curves by the square root of Jeffreys divergence. We consider experimentally the linear interpolation curves in the ordinary, natural and expectation parameterizations of the normal distributions, and compare these curves with a curve derived from the Calvo and Oller's isometric embedding of the Fisher-Rao $d$-variate normal manifold into the cone of $(d+1)\times (d+1)$ symmetric positive-definite matrices [Journal of multivariate analysis 35.2 (1990): 223-242]. We report on our experiments and assess the quality of our approximation technique by comparing the numerical approximations with lower and upper bounds. Finally, we present some information-geometric properties of the Calvo and Oller's isometric embedding.
comment: 30 pages, 16 figures, 3 tables
♻ ☆ Phase-Shifting Coder: Predicting Accurate Orientation in Oriented Object Detection CVPR 2023
With the vigorous development of computer vision, oriented object detection has gradually been featured. In this paper, a novel differentiable angle coder named phase-shifting coder (PSC) is proposed to accurately predict the orientation of objects, along with a dual-frequency version (PSCD). By mapping the rotational periodicity of different cycles into the phase of different frequencies, we provide a unified framework for various periodic fuzzy problems caused by rotational symmetry in oriented object detection. Upon such a framework, common problems in oriented object detection such as boundary discontinuity and square-like problems are elegantly solved in a unified form. Visual analysis and experiments on three datasets prove the effectiveness and the potentiality of our approach. When facing scenarios requiring high-quality bounding boxes, the proposed methods are expected to give a competitive performance. The codes are publicly available at https://github.com/open-mmlab/mmrotate.
comment: Accepted to CVPR 2023, 10 pages, 4 figures
♻ ☆ ABAW: Valence-Arousal Estimation, Expression Recognition, Action Unit Detection & Emotional Reaction Intensity Estimation Challenges
The fifth Affective Behavior Analysis in-the-wild (ABAW) Competition is part of the respective ABAW Workshop which will be held in conjunction with IEEE Computer Vision and Pattern Recognition Conference (CVPR), 2023. The 5th ABAW Competition is a continuation of the Competitions held at ECCV 2022, IEEE CVPR 2022, ICCV 2021, IEEE FG 2020 and CVPR 2017 Conferences, and is dedicated at automatically analyzing affect. For this year's Competition, we feature two corpora: i) an extended version of the Aff-Wild2 database and ii) the Hume-Reaction dataset. The former database is an audiovisual one of around 600 videos of around 3M frames and is annotated with respect to:a) two continuous affect dimensions -valence (how positive/negative a person is) and arousal (how active/passive a person is)-; b) basic expressions (e.g. happiness, sadness, neutral state); and c) atomic facial muscle actions (i.e., action units). The latter dataset is an audiovisual one in which reactions of individuals to emotional stimuli have been annotated with respect to seven emotional expression intensities. Thus the 5th ABAW Competition encompasses four Challenges: i) uni-task Valence-Arousal Estimation, ii) uni-task Expression Classification, iii) uni-task Action Unit Detection, and iv) Emotional Reaction Intensity Estimation. In this paper, we present these Challenges, along with their corpora, we outline the evaluation metrics, we present the baseline systems and illustrate their obtained performance.
comment: arXiv admin note: text overlap with arXiv:2202.10659
♻ ☆ DULDA: Dual-domain Unsupervised Learned Descent Algorithm for PET image reconstruction
Deep learning based PET image reconstruction methods have achieved promising results recently. However, most of these methods follow a supervised learning paradigm, which rely heavily on the availability of high-quality training labels. In particular, the long scanning time required and high radiation exposure associated with PET scans make obtaining this labels impractical. In this paper, we propose a dual-domain unsupervised PET image reconstruction method based on learned decent algorithm, which reconstructs high-quality PET images from sinograms without the need for image labels. Specifically, we unroll the proximal gradient method with a learnable l2,1 norm for PET image reconstruction problem. The training is unsupervised, using measurement domain loss based on deep image prior as well as image domain loss based on rotation equivariance property. The experimental results domonstrate the superior performance of proposed method compared with maximum likelihood expectation maximazation (MLEM), total-variation regularized EM (EM-TV) and deep image prior based method (DIP).
♻ ☆ MIO : Mutual Information Optimization using Self-Supervised Binary Contrastive Learning
Self-supervised contrastive learning frameworks have progressed rapidly over the last few years. In this paper, we propose a novel mutual information optimization-based loss function for contrastive learning. We model our pre-training task as a binary classification problem to induce an implicit contrastive effect and predict whether a pair is positive or negative. We further improve the n\"aive loss function using the Majorize-Minimizer principle and such improvement helps us to track the problem mathematically. Unlike the existing methods, the proposed loss function optimizes the mutual information in both positive and negative pairs. We also present a closed-form expression for the parameter gradient flow and compare the behavior of the proposed loss function using its Hessian eigen-spectrum to analytically study the convergence of SSL frameworks. The proposed method outperforms the SOTA contrastive self-supervised frameworks on benchmark datasets like CIFAR-10, CIFAR-100, STL-10, and Tiny-ImageNet. After 200 epochs of pre-training with ResNet-18 as the backbone, the proposed model achieves an accuracy of 86.2\%, 58.18\%, 77.49\%, and 30.87\% on CIFAR-10, CIFAR-100, STL-10, and Tiny-ImageNet datasets, respectively, and surpasses the SOTA contrastive baseline by 1.23\%, 3.57\%, 2.00\%, and 0.33\%, respectively.
♻ ☆ Cost-Aware Evaluation and Model Scaling for LiDAR-Based 3D Object Detection ICRA 2023
Considerable research effort has been devoted to LiDAR-based 3D object detection and empirical performance has been significantly improved. While progress has been encouraging, we observe an overlooked issue: it is not yet common practice to compare different 3D detectors under the same cost, e.g., inference latency. This makes it difficult to quantify the true performance gain brought by recently proposed architecture designs. The goal of this work is to conduct a cost-aware evaluation of LiDAR-based 3D object detectors. Specifically, we focus on SECOND, a simple grid-based one-stage detector, and analyze its performance under different costs by scaling its original architecture. Then we compare the family of scaled SECOND with recent 3D detection methods, such as Voxel R-CNN and PV-RCNN++. The results are surprising. We find that, if allowed to use the same latency, SECOND can match the performance of PV-RCNN++, the current state-of-the-art method on the Waymo Open Dataset. Scaled SECOND also easily outperforms many recent 3D detection methods published during the past year. We recommend future research control the inference cost in their empirical comparison and include the family of scaled SECOND as a strong baseline when presenting novel 3D detection methods.
comment: ICRA 2023
♻ ☆ DDS3D: Dense Pseudo-Labels with Dynamic Threshold for Semi-Supervised 3D Object Detection ICRA
In this paper, we present a simple yet effective semi-supervised 3D object detector named DDS3D. Our main contributions have two-fold. On the one hand, different from previous works using Non-Maximal Suppression (NMS) or its variants for obtaining the sparse pseudo labels, we propose a dense pseudo-label generation strategy to get dense pseudo-labels, which can retain more potential supervision information for the student network. On the other hand, instead of traditional fixed thresholds, we propose a dynamic threshold manner to generate pseudo-labels, which can guarantee the quality and quantity of pseudo-labels during the whole training process. Benefiting from these two components, our DDS3D outperforms the state-of-the-art semi-supervised 3d object detection with mAP of 3.1% on the pedestrian and 2.1% on the cyclist under the same configuration of 1% samples. Extensive ablation studies on the KITTI dataset demonstrate the effectiveness of our DDS3D. The code and models will be made publicly available at https://github.com/hust-jy/DDS3D
comment: Accepted for publication in 2023 IEEE International Conference on Robotics and Automation (ICRA)
♻ ☆ ParaFormer: Parallel Attention Transformer for Efficient Feature Matching AAAI 2023
Heavy computation is a bottleneck limiting deep-learningbased feature matching algorithms to be applied in many realtime applications. However, existing lightweight networks optimized for Euclidean data cannot address classical feature matching tasks, since sparse keypoint based descriptors are expected to be matched. This paper tackles this problem and proposes two concepts: 1) a novel parallel attention model entitled ParaFormer and 2) a graph based U-Net architecture with attentional pooling. First, ParaFormer fuses features and keypoint positions through the concept of amplitude and phase, and integrates self- and cross-attention in a parallel manner which achieves a win-win performance in terms of accuracy and efficiency. Second, with U-Net architecture and proposed attentional pooling, the ParaFormer-U variant significantly reduces computational complexity, and minimize performance loss caused by downsampling. Sufficient experiments on various applications, including homography estimation, pose estimation, and image matching, demonstrate that ParaFormer achieves state-of-the-art performance while maintaining high efficiency. The efficient ParaFormer-U variant achieves comparable performance with less than 50% FLOPs of the existing attention-based models.
comment: Have been accepted by AAAI 2023
♻ ☆ Implicit Identity Leakage: The Stumbling Block to Improving Deepfake Detection Generalization CVPR 2023
In this paper, we analyse the generalization ability of binary classifiers for the task of deepfake detection. We find that the stumbling block to their generalization is caused by the unexpected learned identity representation on images. Termed as the Implicit Identity Leakage, this phenomenon has been qualitatively and quantitatively verified among various DNNs. Furthermore, based on such understanding, we propose a simple yet effective method named the ID-unaware Deepfake Detection Model to reduce the influence of this phenomenon. Extensive experimental results demonstrate that our method outperforms the state-of-the-art in both in-dataset and cross-dataset evaluation. The code is available at https://github.com/megvii-research/CADDM.
comment: Accepted by CVPR 2023
♻ ☆ EgoTracks: A Long-term Egocentric Visual Object Tracking Dataset
Visual object tracking is a key component to many egocentric vision problems. However, the full spectrum of challenges of egocentric tracking faced by an embodied AI is underrepresented in many existing datasets; these tend to focus on relatively short, third-person videos. Egocentric video has several distinguishing characteristics from those commonly found in past datasets: frequent large camera motions and hand interactions with objects commonly lead to occlusions or objects exiting the frame, and object appearance can change rapidly due to widely different points of view, scale, or object states. Embodied tracking is also naturally long-term, and being able to consistently (re-)associate objects to their appearances and disappearances over as long as a lifetime is critical. Previous datasets under-emphasize this re-detection problem, and their "framed" nature has led to adoption of various spatiotemporal priors that we find do not necessarily generalize to egocentric video. We thus introduce EgoTracks, a new dataset for long-term egocentric visual object tracking. Sourced from the Ego4D dataset, this new dataset presents a significant challenge to recent state-of-the-art single-object tracking models, which we find score poorly on traditional tracking metrics for our new dataset, compared to popular benchmarks. We further show improvements that can be made to a STARK tracker to significantly increase its performance on egocentric data, resulting in a baseline model we call EgoSTARK. We publicly release our annotations and benchmark, hoping our dataset leads to further advancements in tracking.
♻ ☆ Masked Unsupervised Self-training for Label-free Image Classification
State-of-the-art computer vision models are mostly trained with supervised learning using human-labeled images, which limits their scalability due to the expensive annotation cost. While self-supervised representation learning has achieved impressive progress, it still requires a second stage of finetuning on labeled data. On the other hand, models pre-trained with large-scale text-image supervision (e.g., CLIP) have enabled zero-shot transfer to downstream image classification tasks. However, the zero-shot performance of CLIP-like models are often insufficient for real-world adoption. In this paper, we aim to leverage the abundant unlabeled data from a target domain to improve the performance of a pre-trained zero-shot classifier, by unsupervised finetuning of the pre-trained model. We propose Masked Unsupervised Self-Training (MUST), a new unsupervised adaptation method which leverages two different and complementary sources of training signals: pseudo-labels and raw images. MUST jointly optimizes three objectives to learn both class-level global feature and pixel-level local feature and enforces a regularization between the two. We demonstrate the efficacy of MUST on a variety of downstream tasks, where it improves upon CLIP by a large margin. MUST also outperforms supervised few-shot adaptation methods. It achieves a top-1 accuracy of 77.7% on ImageNet using ViT-B, +9.4% higher than CLIP, and +6.2% higher than 16-shot CLIP adaptation. Our code is available at https://github.com/salesforce/MUST.
Information Retrieval
☆ HiNet: A Novel Multi-Scenario & Multi-Task Learning Approach with Hierarchical Information Extraction
Multi-scenario & multi-task learning has been widely applied to many recommendation systems in industrial applications, wherein an effective and practical approach is to carry out multi-scenario transfer learning on the basis of the Mixture-of-Expert (MoE) architecture. However, the MoE-based method, which aims to project all information in the same feature space, cannot effectively deal with the complex relationships inherent among various scenarios and tasks, resulting in unsatisfactory performance. To tackle the problem, we propose a Hierarchical information extraction Network (HiNet) for multi-scenario and multi-task recommendation, which achieves hierarchical extraction based on coarse-to-fine knowledge transfer scheme. The multiple extraction layers of the hierarchical network enable the model to enhance the capability of transferring valuable information across scenarios while preserving specific features of scenarios and tasks. Furthermore, a novel scenario-aware attentive network module is proposed to model correlations between scenarios explicitly. Comprehensive experiments conducted on real-world industrial datasets from Meituan Meishi platform demonstrate that HiNet achieves a new state-of-the-art performance and significantly outperforms existing solutions. HiNet is currently fully deployed in two scenarios and has achieved 2.87% and 1.75% order quantity gain respectively.
☆ Gradient Coordination for Quantifying and Maximizing Knowledge Transference in Multi-Task Learning
Multi-task learning (MTL) has been widely applied in online advertising and recommender systems. To address the negative transfer issue, recent studies have proposed optimization methods that thoroughly focus on the gradient alignment of directions or magnitudes. However, since prior study has proven that both general and specific knowledge exist in the limited shared capacity, overemphasizing on gradient alignment may crowd out task-specific knowledge, and vice versa. In this paper, we propose a transference-driven approach CoGrad that adaptively maximizes knowledge transference via Coordinated Gradient modification. We explicitly quantify the transference as loss reduction from one task to another, and then derive an auxiliary gradient from optimizing it. We perform the optimization by incorporating this gradient into original task gradients, making the model automatically maximize inter-task transfer and minimize individual losses. Thus, CoGrad can harmonize between general and specific knowledge to boost overall performance. Besides, we introduce an efficient approximation of the Hessian matrix, making CoGrad computationally efficient and simple to implement. Both offline and online experiments verify that CoGrad significantly outperforms previous methods.
comment: 5 pages, 3 figures
☆ Semi-supervised Adversarial Learning for Complementary Item Recommendation
Complementary item recommendations are a ubiquitous feature of modern e-commerce sites. Such recommendations are highly effective when they are based on collaborative signals like co-purchase statistics. In certain online marketplaces, however, e.g., on online auction sites, constantly new items are added to the catalog. In such cases, complementary item recommendations are often based on item side-information due to a lack of interaction data. In this work, we propose a novel approach that can leverage both item side-information and labeled complementary item pairs to generate effective complementary recommendations for cold items, i.e., for items for which no co-purchase statistics yet exist. Given that complementary items typically have to be of a different category than the seed item, we technically maintain a latent space for each item category. Simultaneously, we learn to project distributed item representations into these category spaces to determine suitable recommendations. The main learning process in our architecture utilizes labeled pairs of complementary items. In addition, we adopt ideas from Cycle Generative Adversarial Networks (CycleGAN) to leverage available item information even in case no labeled data exists for a given item and category. Experiments on three e-commerce datasets show that our method is highly effective.
comment: ACM Web Conference 2023
☆ Pacos: Modeling Users' Interpretable and Context-Dependent Choices in Preference Reversals
Choice problems refer to selecting the best choices from several items, and learning users' preferences in choice problems is of great significance in understanding the decision making mechanisms and providing personalized services. Existing works typically assume that people evaluate items independently. In practice, however, users' preferences depend on the market in which items are placed, which is known as context effects; and the order of users' preferences for two items may even be reversed, which is referred to preference reversals. In this work, we identify three factors contributing to context effects: users' adaptive weights, the inter-item comparison, and display positions. We propose a context-dependent preference model named Pacos as a unified framework for addressing three factors simultaneously, and consider two design methods including an additive method with high interpretability and an ANN-based method with high accuracy. We study the conditions for preference reversals to occur and provide an theoretical proof of the effectiveness of Pacos in addressing preference reversals. Experimental results show that the proposed method has better performance than prior works in predicting users' choices, and has great interpretability to help understand the cause of preference reversals.
comment: 29 pages, 12 figures
♻ ☆ Multi-Task Recommendations with Reinforcement Learning
In recent years, Multi-task Learning (MTL) has yielded immense success in Recommender System (RS) applications. However, current MTL-based recommendation models tend to disregard the session-wise patterns of user-item interactions because they are predominantly constructed based on item-wise datasets. Moreover, balancing multiple objectives has always been a challenge in this field, which is typically avoided via linear estimations in existing works. To address these issues, in this paper, we propose a Reinforcement Learning (RL) enhanced MTL framework, namely RMTL, to combine the losses of different recommendation tasks using dynamic weights. To be specific, the RMTL structure can address the two aforementioned issues by (i) constructing an MTL environment from session-wise interactions and (ii) training multi-task actor-critic network structure, which is compatible with most existing MTL-based recommendation models, and (iii) optimizing and fine-tuning the MTL loss function using the weights generated by critic networks. Experiments on two real-world public datasets demonstrate the effectiveness of RMTL with a higher AUC against state-of-the-art MTL-based recommendation models. Additionally, we evaluate and validate RMTL's compatibility and transferability across various MTL models.
comment: TheWebConf2023
♻ ☆ Achievable Rates and Low-Complexity Encoding of Posterior Matching for the BSC
Horstein, Burnashev, Shayevitz and Feder, Naghshvar et al. and others have studied sequential transmission of a K-bit message over the binary symmetric channel (BSC) with full, noiseless feedback using posterior matching. Yang et al. provide an improved lower bound on the achievable rate using martingale analysis that relies on the small-enough difference (SED) partitioning introduced by Naghshvar et al. SED requires a relatively complex encoder and decoder. To reduce complexity, this paper replaces SED with relaxed constraints that admit the small enough absolute difference (SEAD) partitioning rule. The main analytical results show that achievable-rate bounds higher than those found by Yang et al. are possible even under the new constraints, which are less restrictive than SED. The new analysis does not use martingale theory for the confirmation phase and applies a surrogate channel technique to tighten the results. An initial systematic transmission further increases the achievable rate bound. The simplified encoder associated with SEAD has a complexity below order O(K^2) and allows simulations for message sizes of at least 1000 bits. For example, simulations achieve 99% of of the channel's 0.50-bit capacity with an average block size of 200 bits for a target codeword error rate of 10^(-3).
comment: This paper consists of 26 pages and contains 6 figures. An earlier version of the algorithm included in this paper was published at the 2020 IEEE International Symposium on Information Theory (ISIT), (DOI: 10.1109/ISIT44484.2020.9174232)
Machine Learning
☆ Exphormer: Sparse Transformers for Graphs
Graph transformers have emerged as a promising architecture for a variety of graph learning and representation tasks. Despite their successes, though, it remains challenging to scale graph transformers to large graphs while maintaining accuracy competitive with message-passing networks. In this paper, we introduce Exphormer, a framework for building powerful and scalable graph transformers. Exphormer consists of a sparse attention mechanism based on two mechanisms: virtual global nodes and expander graphs, whose mathematical characteristics, such as spectral expansion, pseduorandomness, and sparsity, yield graph transformers with complexity only linear in the size of the graph, while allowing us to prove desirable theoretical properties of the resulting transformer models. We show that incorporating \textsc{Exphormer} into the recently-proposed GraphGPS framework produces models with competitive empirical results on a wide variety of graph datasets, including state-of-the-art results on three datasets. We also show that \textsc{Exphormer} can scale to datasets on larger graphs than shown in previous graph transformer architectures. Code can be found at https://github.com/hamed1375/Exphormer.
☆ StyleGANEX: StyleGAN-Based Manipulation Beyond Cropped Aligned Faces
Recent advances in face manipulation using StyleGAN have produced impressive results. However, StyleGAN is inherently limited to cropped aligned faces at a fixed image resolution it is pre-trained on. In this paper, we propose a simple and effective solution to this limitation by using dilated convolutions to rescale the receptive fields of shallow layers in StyleGAN, without altering any model parameters. This allows fixed-size small features at shallow layers to be extended into larger ones that can accommodate variable resolutions, making them more robust in characterizing unaligned faces. To enable real face inversion and manipulation, we introduce a corresponding encoder that provides the first-layer feature of the extended StyleGAN in addition to the latent style code. We validate the effectiveness of our method using unaligned face inputs of various resolutions in a diverse set of face manipulation tasks, including facial attribute editing, super-resolution, sketch/mask-to-face translation, and face toonification.
comment: Code: https://github.com/williamyang1991/StyleGANEX Project page: https://www.mmlab-ntu.com/project/styleganex/
☆ Multiple Hands Make Light Work: Enhancing Quality and Diversity using MAP-Elites with Multiple Parallel Evolution Strategies
With the development of hardware accelerators and their corresponding tools, evaluations have become more affordable through fast and massively parallel evaluations in some applications. This advancement has drastically sped up the runtime of evolution-inspired algorithms such as Quality-Diversity optimization, creating tremendous potential for algorithmic innovation through scale. In this work, we propose MAP-Elites-Multi-ES (MEMES), a novel QD algorithm based on Evolution Strategies (ES) designed for fast parallel evaluations. ME-Multi-ES builds on top of the existing MAP-Elites-ES algorithm, scaling it by maintaining multiple independent ES threads with massive parallelization. We also introduce a new dynamic reset procedure for the lifespan of the independent ES to autonomously maximize the improvement of the QD population. We show experimentally that MEMES outperforms existing gradient-based and objective-agnostic QD algorithms when compared in terms of generations. We perform this comparison on both black-box optimization and QD-Reinforcement Learning tasks, demonstrating the benefit of our approach across different problems and domains. Finally, we also find that our approach intrinsically enables optimization of fitness locally around a niche, a phenomenon not observed in other QD algorithms.
☆ Rewarding Chatbots for Real-World Engagement with Millions of Users
The emergence of pretrained large language models has led to the deployment of a range of social chatbots for chitchat. Although these chatbots demonstrate language ability and fluency, they are not guaranteed to be engaging and can struggle to retain users. This work investigates the development of social chatbots that prioritize user engagement to enhance retention, specifically examining the use of human feedback to efficiently develop highly engaging chatbots. The proposed approach uses automatic pseudo-labels collected from user interactions to train a reward model that can be used to reject low-scoring sample responses generated by the chatbot model at inference time. Intuitive evaluation metrics, such as mean conversation length (MCL), are introduced as proxies to measure the level of engagement of deployed chatbots. A/B testing on groups of 10,000 new daily chatbot users on the Chai Research platform shows that this approach increases the MCL by up to 70%, which translates to a more than 30% increase in user retention for a GPT-J 6B model. Future work aims to use the reward model to realise a data fly-wheel, where the latest user conversations can be used to alternately fine-tune the language model and the reward model.
☆ Ignorance is Bliss: Robust Control via Information Gating
Informational parsimony -- i.e., using the minimal information required for a task, -- provides a useful inductive bias for learning representations that achieve better generalization by being robust to noise and spurious correlations. We propose information gating in the pixel space as a way to learn more parsimonious representations. Information gating works by learning masks that capture only the minimal information required to solve a given task. Intuitively, our models learn to identify which visual cues actually matter for a given task. We gate information using a differentiable parameterization of the signal-to-noise ratio, which can be applied to arbitrary values in a network, e.g.~masking out pixels at the input layer. We apply our approach, which we call InfoGating, to various objectives such as: multi-step forward and inverse dynamics, Q-learning, behavior cloning, and standard self-supervised tasks. Our experiments show that learning to identify and use minimal information can improve generalization in downstream tasks -- e.g., policies based on info-gated images are considerably more robust to distracting/irrelevant visual features.
☆ On the Fusion Strategies for Federated Decision Making
We consider the problem of information aggregation in federated decision making, where a group of agents collaborate to infer the underlying state of nature without sharing their private data with the central processor or each other. We analyze the non-Bayesian social learning strategy in which agents incorporate their individual observations into their opinions (i.e., soft-decisions) with Bayes rule, and the central processor aggregates these opinions by arithmetic or geometric averaging. Building on our previous work, we establish that both pooling strategies result in asymptotic normality characterization of the system, which, for instance, can be utilized in order to give approximate expressions for the error probability. We verify the theoretical findings with simulations and compare both strategies.
comment: Submitted for publication
☆ Long-tailed Classification from a Bayesian-decision-theory Perspective
Long-tailed classification poses a challenge due to its heavy imbalance in class probabilities and tail-sensitivity risks with asymmetric misprediction costs. Recent attempts have used re-balancing loss and ensemble methods, but they are largely heuristic and depend heavily on empirical results, lacking theoretical explanation. Furthermore, existing methods overlook the decision loss, which characterizes different costs associated with tailed classes. This paper presents a general and principled framework from a Bayesian-decision-theory perspective, which unifies existing techniques including re-balancing and ensemble methods, and provides theoretical justifications for their effectiveness. From this perspective, we derive a novel objective based on the integrated risk and a Bayesian deep-ensemble approach to improve the accuracy of all classes, especially the ``tail". Besides, our framework allows for task-adaptive decision loss which provides provably optimal decisions in varying task scenarios, along with the capability to quantify uncertainty. Finally, We conduct comprehensive experiments, including standard classification, tail-sensitive classification with a new False Head Rate metric, calibration, and ablation studies. Our framework significantly improves the current SOTA even on large-scale real-world datasets like ImageNet.
☆ Modeling Events and Interactions through Temporal Processes -- A Survey
In real-world scenario, many phenomena produce a collection of events that occur in continuous time. Point Processes provide a natural mathematical framework for modeling these sequences of events. In this survey, we investigate probabilistic models for modeling event sequences through temporal processes. We revise the notion of event modeling and provide the mathematical foundations that characterize the literature on the topic. We define an ontology to categorize the existing approaches in terms of three families: simple, marked, and spatio-temporal point processes. For each family, we systematically review the existing approaches based based on deep learning. Finally, we analyze the scenarios where the proposed techniques can be used for addressing prediction and modeling aspects.
☆ A General Recipe for the Analysis of Randomized Multi-Armed Bandit Algorithms
In this paper we propose a general methodology to derive regret bounds for randomized multi-armed bandit algorithms. It consists in checking a set of sufficient conditions on the sampling probability of each arm and on the family of distributions to prove a logarithmic regret. As a direct application we revisit two famous bandit algorithms, Minimum Empirical Divergence (MED) and Thompson Sampling (TS), under various models for the distributions including single parameter exponential families, Gaussian distributions, bounded distributions, or distributions satisfying some conditions on their moments. In particular, we prove that MED is asymptotically optimal for all these models, but also provide a simple regret analysis of some TS algorithms for which the optimality is already known. We then further illustrate the interest of our approach, by analyzing a new Non-Parametric TS algorithm (h-NPTS), adapted to some families of unbounded reward distributions with a bounded h-moment. This model can for instance capture some non-parametric families of distributions whose variance is upper bounded by a known constant.
☆ TSMixer: An all-MLP Architecture for Time Series Forecasting
Real-world time-series datasets are often multivariate with complex dynamics. Commonly-used high capacity architectures like recurrent- or attention-based sequential models have become popular. However, recent work demonstrates that simple univariate linear models can outperform those deep alternatives. In this paper, we investigate the capabilities of linear models for time-series forecasting and present Time-Series Mixer (TSMixer), an architecture designed by stacking multi-layer perceptrons (MLPs). TSMixer is based on mixing operations along time and feature dimensions to extract information efficiently. On popular academic benchmarks, the simple-to-implement TSMixer is comparable to specialized state-of-the-art models that leverage the inductive biases of specific benchmarks. On the challenging and large scale M5 benchmark, a real-world retail dataset, TSMixer demonstrates superior performance compared to the state-of-the-art alternatives. Our results underline the importance of efficiently utilizing cross-variate and auxiliary information for improving the performance of time series forecasting. The design paradigms utilized in TSMixer are expected to open new horizons for deep learning-based time series forecasting.
☆ Tactile-Filter: Interactive Tactile Perception for Part Mating
Humans rely on touch and tactile sensing for a lot of dexterous manipulation tasks. Our tactile sensing provides us with a lot of information regarding contact formations as well as geometric information about objects during any interaction. With this motivation, vision-based tactile sensors are being widely used for various robotic perception and control tasks. In this paper, we present a method for interactive perception using vision-based tactile sensors for multi-object assembly. In particular, we are interested in tactile perception during part mating, where a robot can use tactile sensors and a feedback mechanism using particle filter to incrementally improve its estimate of objects that fit together for assembly. To do this, we first train a deep neural network that makes use of tactile images to predict the probabilistic correspondence between arbitrarily shaped objects that fit together. The trained model is used to design a particle filter which is used twofold. First, given one partial (or non-unique) observation of the hole, it incrementally improves the estimate of the correct peg by sampling more tactile observations. Second, it selects the next action for the robot to sample the next touch (and thus image) which results in maximum uncertainty reduction to minimize the number of interactions during the perception task. We evaluate our method on several part-mating tasks for assembly using a robot equipped with a vision-based tactile sensor. We also show the efficiency of the proposed action selection method against a naive method. See supplementary video at https://www.youtube.com/watch?v=jMVBg_e3gLw .
comment: under submission
☆ A hybrid deep-learning-metaheuristic framework to approximate discrete road network design problems
This study proposes a hybrid deep-learning-metaheuristic framework with a bi-level architecture to solve road network design problems (NDPs). We train a graph neural network (GNN) to approximate the solution of the user equilibrium (UE) traffic assignment problem, and use inferences made by the trained model to calculate fitness function evaluations of a genetic algorithm (GA) to approximate solutions for NDPs. Using two NDP variants and an exact solver as benchmark, we show that our proposed framework can provide solutions within 5% gap of the global optimum results given less than 1% of the time required for finding the optimal results. Moreover, we observe many interesting future directions, thus we propose a brief research agenda for this topic. The key observation inspiring influential future research was that fitness function evaluation time using the inferences made by the GNN model for the genetic algorithm was in the order of milliseconds, which points to an opportunity and a need for novel heuristics that 1) can cope well with noisy fitness function values provided by neural networks, and 2) can use the significantly higher computation time provided to them to explore the search space effectively (rather than efficiently). This opens a new avenue for a modern class of metaheuristics that are crafted for use with AI-powered predictors.
☆ Machine learning for sports betting: should forecasting models be optimised for accuracy or calibration?
Sports betting's recent federal legalisation in the USA coincides with the golden age of machine learning. If bettors can leverage data to accurately predict the probability of an outcome, they can recognise when the bookmaker's odds are in their favour. As sports betting is a multi-billion dollar industry in the USA alone, identifying such opportunities could be extremely lucrative. Many researchers have applied machine learning to the sports outcome prediction problem, generally using accuracy to evaluate the performance of forecasting models. We hypothesise that for the sports betting problem, model calibration is more important than accuracy. To test this hypothesis, we train models on NBA data over several seasons and run betting experiments on a single season, using published odds. Evaluating various betting systems, we show that optimising the forecasting model for calibration leads to greater returns than optimising for accuracy, on average (return on investment of $110.42\%$ versus $2.98\%$) and in the best case ($902.01\%$ versus $222.84\%$). These findings suggest that for sports betting (or any forecasting problem where decisions are made based on the predicted probability of each outcome), calibration is a more important metric than accuracy. Sports bettors who wish to increase profits should therefore optimise their forecasting model for calibration.
comment: 27 pages (including bibliography). 8 Figures
☆ Forecasting Solar Irradiance without Direct Observation: An Empirical Analysis
As the use of solar power increases, having accurate and timely forecasters will be essential for smooth grid operators. There are many proposed methods for forecasting solar irradiance / solar power production. However, many of these methods formulate the problem as a time-series, relying on near real-time access to observations at the location of interest to generate forecasts. This requires both access to a real-time stream of data and enough historical observations for these methods to be deployed. In this paper, we conduct a thorough analysis of effective ways to formulate the forecasting problem comparing classical machine learning approaches to state-of-the-art deep learning. Using data from 20 locations distributed throughout the UK and commercially available weather data, we show that it is possible to build systems that do not require access to this data. Leveraging weather observations and measurements from other locations we show it is possible to create models capable of accurately forecasting solar irradiance at new locations. We utilise compare both satellite and ground observations (e.g. temperature, pressure) of weather data. This could facilitate use planning and optimisation for both newly deployed solar farms and domestic installations from the moment they come online. Additionally, we show that training a single global model for multiple locations can produce a more robust model with more consistent and accurate results across locations.
comment: 39 pages, 11 figures
☆ Neural Gromov-Wasserstein Optimal Transport
We present a scalable neural method to solve the Gromov-Wasserstein (GW) Optimal Transport (OT) problem with the inner product cost. In this problem, given two distributions supported on (possibly different) spaces, one has to find the most isometric map between them. Our proposed approach uses neural networks and stochastic mini-batch optimization which allows to overcome the limitations of existing GW methods such as their poor scalability with the number of samples and the lack of out-of-sample estimation. To demonstrate the effectiveness of our proposed method, we conduct experiments on the synthetic data and explore the practical applicability of our method to the popular task of the unsupervised alignment of word embeddings.
☆ Classifying the evolution of COVID-19 severity on patients with combined dynamic Bayesian networks and neural networks
When we face patients arriving to a hospital suffering from the effects of some illness, one of the main problems we can encounter is evaluating whether or not said patients are going to require intensive care in the near future. This intensive care requires allotting valuable and scarce resources, and knowing beforehand the severity of a patients illness can improve both its treatment and the organization of resources. We illustrate this issue in a dataset consistent of Spanish COVID-19 patients from the sixth epidemic wave where we label patients as critical when they either had to enter the intensive care unit or passed away. We then combine the use of dynamic Bayesian networks, to forecast the vital signs and the blood analysis results of patients over the next 40 hours, and neural networks, to evaluate the severity of a patients disease in that interval of time. Our empirical results show that the transposition of the current state of a patient to future values with the DBN for its subsequent use in classification obtains better the accuracy and g-mean score than a direct application with a classifier.
☆ Understanding and Constructing Latent Modality Structures in Multi-modal Representation Learning CVPR 2023
Contrastive loss has been increasingly used in learning representations from multiple modalities. In the limit, the nature of the contrastive loss encourages modalities to exactly match each other in the latent space. Yet it remains an open question how the modality alignment affects the downstream task performance. In this paper, based on an information-theoretic argument, we first prove that exact modality alignment is sub-optimal in general for downstream prediction tasks. Hence we advocate that the key of better performance lies in meaningful latent modality structures instead of perfect modality alignment. To this end, we propose three general approaches to construct latent modality structures. Specifically, we design 1) a deep feature separation loss for intra-modality regularization; 2) a Brownian-bridge loss for inter-modality regularization; and 3) a geometric consistency loss for both intra- and inter-modality regularization. Extensive experiments are conducted on two popular multi-modal representation learning frameworks: the CLIP-based two-tower model and the ALBEF-based fusion model. We test our model on a variety of tasks including zero/few-shot image classification, image-text retrieval, visual question answering, visual reasoning, and visual entailment. Our method achieves consistent improvements over existing methods, demonstrating the effectiveness and generalizability of our proposed approach on latent modality structure regularization.
comment: 14 pages, 8 figure, CVPR 2023 accepted
☆ Automotive Perception Software Development: An Empirical Investigation into Data, Annotation, and Ecosystem Challenges
Software that contains machine learning algorithms is an integral part of automotive perception, for example, in driving automation systems. The development of such software, specifically the training and validation of the machine learning components, require large annotated datasets. An industry of data and annotation services has emerged to serve the development of such data-intensive automotive software components. Wide-spread difficulties to specify data and annotation needs challenge collaborations between OEMs (Original Equipment Manufacturers) and their suppliers of software components, data, and annotations. This paper investigates the reasons for these difficulties for practitioners in the Swedish automotive industry to arrive at clear specifications for data and annotations. The results from an interview study show that a lack of effective metrics for data quality aspects, ambiguities in the way of working, unclear definitions of annotation quality, and deficits in the business ecosystems are causes for the difficulty in deriving the specifications. We provide a list of recommendations that can mitigate challenges when deriving specifications and we propose future research opportunities to overcome these challenges. Our work contributes towards the on-going research on accountability of machine learning as applied to complex software systems, especially for high-stake applications such as automated driving.
☆ Estimating friction coefficient using generative modelling
It is common to utilise dynamic models to measure the tyre-road friction in real-time. Alternatively, predictive approaches estimate the tyre-road friction by identifying the environmental factors affecting it. This work aims to formulate the problem of friction estimation as a visual perceptual learning task. The problem is broken down into detecting surface characteristics by applying semantic segmentation and using the extracted features to predict the frictional force. This work for the first time formulates the friction estimation problem as a regression from the latent space of a semantic segmentation model. The preliminary results indicate that this approach can estimate frictional force.
comment: To be published in ICM2023
☆ Lifelong Machine Learning Potentials
Machine learning potentials (MLPs) trained on accurate quantum chemical data can retain the high accuracy, while inflicting little computational demands. On the downside, they need to be trained for each individual system. In recent years, a vast number of MLPs has been trained from scratch because learning additional data typically requires to train again on all data to not forget previously acquired knowledge. Additionally, most common structural descriptors of MLPs cannot represent efficiently a large number of different chemical elements. In this work, we tackle these problems by introducing element-embracing atom-centered symmetry functions (eeACSFs) which combine structural properties and element information from the periodic table. These eeACSFs are a key for our development of a lifelong machine learning potential (lMLP). Uncertainty quantification can be exploited to transgress a fixed, pre-trained MLP to arrive at a continuously adapting lMLP, because a predefined level of accuracy can be ensured. To extend the applicability of an lMLP to new systems, we apply continual learning strategies to enable autonomous and on-the-fly training on a continuous stream of new data. For the training of deep neural networks, we propose the continual resilient (CoRe) optimizer and incremental learning strategies relying on rehearsal of data, regularization of parameters, and the architecture of the model.
comment: 20 pages, 6 figures
☆ Product Jacobi-Theta Boltzmann machines with score matching
The estimation of probability density functions is a non trivial task that over the last years has been tackled with machine learning techniques. Successful applications can be obtained using models inspired by the Boltzmann machine (BM) architecture. In this manuscript, the product Jacobi-Theta Boltzmann machine (pJTBM) is introduced as a restricted version of the Riemann-Theta Boltzmann machine (RTBM) with diagonal hidden sector connection matrix. We show that score matching, based on the Fisher divergence, can be used to fit probability densities with the pJTBM more efficiently than with the original RTBM.
comment: 7 pages, 3 figures, ACAT22 proceedings
☆ Deep Anomaly Detection on Tennessee Eastman Process Data
This paper provides the first comprehensive evaluation and analysis of modern (deep-learning) unsupervised anomaly detection methods for chemical process data. We focus on the Tennessee Eastman process dataset, which has been a standard litmus test to benchmark anomaly detection methods for nearly three decades. Our extensive study will facilitate choosing appropriate anomaly detection methods in industrial applications.
☆ Distribution Preserving Source Separation With Time Frequency Predictive Models
We provide an example of a distribution preserving source separation method, which aims at addressing perceptual shortcomings of state-of-the-art methods. Our approach uses unconditioned generative models of signal sources. Reconstruction is achieved by means of mix-consistent sampling from a distribution conditioned on a realization of a mix. The separated signals follow their respective source distributions, which provides an advantage when separation results are evaluated in a listening test.
comment: 5 pages, 4 figures, pre-review version submitted to EUSIPCO 2023
☆ Simulation-based Bayesian inference for robotic grasping IROS 2022
General robotic grippers are challenging to control because of their rich nonsmooth contact dynamics and the many sources of uncertainties due to the environment or sensor noise. In this work, we demonstrate how to compute 6-DoF grasp poses using simulation-based Bayesian inference through the full stochastic forward simulation of the robot in its environment while robustly accounting for many of the uncertainties in the system. A Riemannian manifold optimization procedure preserving the nonlinearity of the rotation space is used to compute the maximum a posteriori grasp pose. Simulation and physical benchmarks show the promising high success rate of the approach.
comment: 5 pages, 4 figures, IROS 2022 Probabilistic Robotics at the age of Deep Learning workshop. arXiv admin note: substantial text overlap with arXiv:2109.14275
☆ Accurate Real-time Polyp Detection in Videos from Concatenation of Latent Features Extracted from Consecutive Frames
An efficient deep learning model that can be implemented in real-time for polyp detection is crucial to reducing polyp miss-rate during screening procedures. Convolutional neural networks (CNNs) are vulnerable to small changes in the input image. A CNN-based model may miss the same polyp appearing in a series of consecutive frames and produce unsubtle detection output due to changes in camera pose, lighting condition, light reflection, etc. In this study, we attempt to tackle this problem by integrating temporal information among neighboring frames. We propose an efficient feature concatenation method for a CNN-based encoder-decoder model without adding complexity to the model. The proposed method incorporates extracted feature maps of previous frames to detect polyps in the current frame. The experimental results demonstrate that the proposed method of feature concatenation improves the overall performance of automatic polyp detection in videos. The following results are obtained on a public video dataset: sensitivity 90.94\%, precision 90.53\%, and specificity 92.46%
☆ Variational Quantum Neural Networks (VQNNS) in Image Classification
Quantum machine learning has established as an interdisciplinary field to overcome limitations of classical machine learning and neural networks. This is a field of research which can prove that quantum computers are able to solve problems with complex correlations between inputs that can be hard for classical computers. This suggests that learning models made on quantum computers may be more powerful for applications, potentially faster computation and better generalization on less data. The objective of this paper is to investigate how training of quantum neural network (QNNs) can be done using quantum optimization algorithms for improving the performance and time complexity of QNNs. A classical neural network can be partially quantized to create a hybrid quantum-classical neural network which is used mainly in classification and image recognition. In this paper, a QNN structure is made where a variational parameterized circuit is incorporated as an input layer named as Variational Quantum Neural Network (VQNNs). We encode the cost function of QNNs onto relative phases of a superposition state in the Hilbert space of the network parameters. The parameters are tuned with an iterative quantum approximate optimisation (QAOA) mixer and problem hamiltonians. VQNNs is experimented with MNIST digit recognition (less complex) and crack image classification datasets (more complex) which converges the computation in lesser time than QNN with decent training accuracy.
☆ Decision-Making Under Uncertainty: Beyond Probabilities
This position paper reflects on the state-of-the-art in decision-making under uncertainty. A classical assumption is that probabilities can sufficiently capture all uncertainty in a system. In this paper, the focus is on the uncertainty that goes beyond this classical interpretation, particularly by employing a clear distinction between aleatoric and epistemic uncertainty. The paper features an overview of Markov decision processes (MDPs) and extensions to account for partial observability and adversarial behavior. These models sufficiently capture aleatoric uncertainty but fail to account for epistemic uncertainty robustly. Consequently, we present a thorough overview of so-called uncertainty models that exhibit uncertainty in a more robust interpretation. We show several solution techniques for both discrete and continuous models, ranging from formal verification, over control-based abstractions, to reinforcement learning. As an integral part of this paper, we list and discuss several key challenges that arise when dealing with rich types of uncertainty in a model-based fashion.
☆ Gradient Coordination for Quantifying and Maximizing Knowledge Transference in Multi-Task Learning
Multi-task learning (MTL) has been widely applied in online advertising and recommender systems. To address the negative transfer issue, recent studies have proposed optimization methods that thoroughly focus on the gradient alignment of directions or magnitudes. However, since prior study has proven that both general and specific knowledge exist in the limited shared capacity, overemphasizing on gradient alignment may crowd out task-specific knowledge, and vice versa. In this paper, we propose a transference-driven approach CoGrad that adaptively maximizes knowledge transference via Coordinated Gradient modification. We explicitly quantify the transference as loss reduction from one task to another, and then derive an auxiliary gradient from optimizing it. We perform the optimization by incorporating this gradient into original task gradients, making the model automatically maximize inter-task transfer and minimize individual losses. Thus, CoGrad can harmonize between general and specific knowledge to boost overall performance. Besides, we introduce an efficient approximation of the Hessian matrix, making CoGrad computationally efficient and simple to implement. Both offline and online experiments verify that CoGrad significantly outperforms previous methods.
comment: 5 pages, 3 figures
☆ Contrastive Language-Image Pretrained (CLIP) Models are Powerful Out-of-Distribution Detectors
We present a comprehensive experimental study on pretrained feature extractors for visual out-of-distribution (OOD) detection. We examine several setups, based on the availability of labels or image captions and using different combinations of in- and out-distributions. Intriguingly, we find that (i) contrastive language-image pretrained models achieve state-of-the-art unsupervised out-of-distribution performance using nearest neighbors feature similarity as the OOD detection score, (ii) supervised state-of-the-art OOD detection performance can be obtained without in-distribution fine-tuning, (iii) even top-performing billion-scale vision transformers trained with natural language supervision fail at detecting adversarially manipulated OOD images. Finally, we argue whether new benchmarks for visual anomaly detection are needed based on our experiments. Using the largest publicly available vision transformer, we achieve state-of-the-art performance across all $18$ reported OOD benchmarks, including an AUROC of 87.6\% (9.2\% gain, unsupervised) and 97.4\% (1.2\% gain, supervised) for the challenging task of CIFAR100 $\rightarrow$ CIFAR10 OOD detection. The code will be open-sourced.
☆ Semi-supervised Adversarial Learning for Complementary Item Recommendation
Complementary item recommendations are a ubiquitous feature of modern e-commerce sites. Such recommendations are highly effective when they are based on collaborative signals like co-purchase statistics. In certain online marketplaces, however, e.g., on online auction sites, constantly new items are added to the catalog. In such cases, complementary item recommendations are often based on item side-information due to a lack of interaction data. In this work, we propose a novel approach that can leverage both item side-information and labeled complementary item pairs to generate effective complementary recommendations for cold items, i.e., for items for which no co-purchase statistics yet exist. Given that complementary items typically have to be of a different category than the seed item, we technically maintain a latent space for each item category. Simultaneously, we learn to project distributed item representations into these category spaces to determine suitable recommendations. The main learning process in our architecture utilizes labeled pairs of complementary items. In addition, we adopt ideas from Cycle Generative Adversarial Networks (CycleGAN) to leverage available item information even in case no labeled data exists for a given item and category. Experiments on three e-commerce datasets show that our method is highly effective.
comment: ACM Web Conference 2023
☆ Distributionally Robust Optimization with Probabilistic Group AAAI 2023
Modern machine learning models may be susceptible to learning spurious correlations that hold on average but not for the atypical group of samples. To address the problem, previous approaches minimize the empirical worst-group risk. Despite the promise, they often assume that each sample belongs to one and only one group, which does not allow expressing the uncertainty in group labeling. In this paper, we propose a novel framework PG-DRO, which explores the idea of probabilistic group membership for distributionally robust optimization. Key to our framework, we consider soft group membership instead of hard group annotations. The group probabilities can be flexibly generated using either supervised learning or zero-shot approaches. Our framework accommodates samples with group membership ambiguity, offering stronger flexibility and generality than the prior art. We comprehensively evaluate PG-DRO on both image classification and natural language processing benchmarks, establishing superior performance
comment: Published at AAAI 2023
☆ Sliced-Wasserstein on Symmetric Positive Definite Matrices for M/EEG Signals
When dealing with electro or magnetoencephalography records, many supervised prediction tasks are solved by working with covariance matrices to summarize the signals. Learning with these matrices requires using Riemanian geometry to account for their structure. In this paper, we propose a new method to deal with distributions of covariance matrices and demonstrate its computational efficiency on M/EEG multivariate time series. More specifically, we define a Sliced-Wasserstein distance between measures of symmetric positive definite matrices that comes with strong theoretical guarantees. Then, we take advantage of its properties and kernel methods to apply this distance to brain-age prediction from MEG data and compare it to state-of-the-art algorithms based on Riemannian geometry. Finally, we show that it is an efficient surrogate to the Wasserstein distance in domain adaptation for Brain Computer Interface applications.
☆ Training, Architecture, and Prior for Deterministic Uncertainty Methods
Accurate and efficient uncertainty estimation is crucial to build reliable Machine Learning (ML) models capable to provide calibrated uncertainty estimates, generalize and detect Out-Of-Distribution (OOD) datasets. To this end, Deterministic Uncertainty Methods (DUMs) is a promising model family capable to perform uncertainty estimation in a single forward pass. This work investigates important design choices in DUMs: (1) we show that training schemes decoupling the core architecture and the uncertainty head schemes can significantly improve uncertainty performances. (2) we demonstrate that the core architecture expressiveness is crucial for uncertainty performance and that additional architecture constraints to avoid feature collapse can deteriorate the trade-off between OOD generalization and detection. (3) Contrary to other Bayesian models, we show that the prior defined by DUMs do not have a strong effect on the final performances.
☆ Deep Generative Fixed-filter Active Noise Control ICASSP 2023
Due to the slow convergence and poor tracking ability, conventional LMS-based adaptive algorithms are less capable of handling dynamic noises. Selective fixed-filter active noise control (SFANC) can significantly reduce response time by selecting appropriate pre-trained control filters for different noises. Nonetheless, the limited number of pre-trained control filters may affect noise reduction performance, especially when the incoming noise differs much from the initial noises during pre-training. Therefore, a generative fixed-filter active noise control (GFANC) method is proposed in this paper to overcome the limitation. Based on deep learning and a perfect-reconstruction filter bank, the GFANC method only requires a few prior data (one pre-trained broadband control filter) to automatically generate suitable control filters for various noises. The efficacy of the GFANC method is demonstrated by numerical simulations on real-recorded noises.
comment: Accepted by ICASSP 2023. Code will be available after publication
☆ Scaling Up 3D Kernels with Bayesian Frequency Re-parameterization for Medical Image Segmentation
With the inspiration of vision transformers, the concept of depth-wise convolution revisits to provide a large Effective Receptive Field (ERF) using Large Kernel (LK) sizes for medical image segmentation. However, the segmentation performance might be saturated and even degraded as the kernel sizes scaled up (e.g., $21\times 21\times 21$) in a Convolutional Neural Network (CNN). We hypothesize that convolution with LK sizes is limited to maintain an optimal convergence for locality learning. While Structural Re-parameterization (SR) enhances the local convergence with small kernels in parallel, optimal small kernel branches may hinder the computational efficiency for training. In this work, we propose RepUX-Net, a pure CNN architecture with a simple large kernel block design, which competes favorably with current network state-of-the-art (SOTA) (e.g., 3D UX-Net, SwinUNETR) using 6 challenging public datasets. We derive an equivalency between kernel re-parameterization and the branch-wise variation in kernel convergence. Inspired by the spatial frequency in the human visual system, we extend to vary the kernel convergence into element-wise setting and model the spatial frequency as a Bayesian prior to re-parameterize convolutional weights during training. Specifically, a reciprocal function is leveraged to estimate a frequency-weighted value, which rescales the corresponding kernel element for stochastic gradient descent. From the experimental results, RepUX-Net consistently outperforms 3D SOTA benchmarks with internal validation (FLARE: 0.929 to 0.944), external validation (MSD: 0.901 to 0.932, KiTS: 0.815 to 0.847, LiTS: 0.933 to 0.949, TCIA: 0.736 to 0.779) and transfer learning (AMOS: 0.880 to 0.911) scenarios in Dice Score.
comment: Both codes and pretrained models are available at: https://github.com/MASILab/RepUX-Net
Self-Supervised CSF Inpainting with Synthetic Atrophy for Improved Accuracy Validation of Cortical Surface Analyses
Accuracy validation of cortical thickness measurement is a difficult problem due to the lack of ground truth data. To address this need, many methods have been developed to synthetically induce gray matter (GM) atrophy in an MRI via deformable registration, creating a set of images with known changes in cortical thickness. However, these methods often cause blurring in atrophied regions, and cannot simulate realistic atrophy within deep sulci where cerebrospinal fluid (CSF) is obscured or absent. In this paper, we present a solution using a self-supervised inpainting model to generate CSF in these regions and create images with more plausible GM/CSF boundaries. Specifically, we introduce a novel, 3D GAN model that incorporates patch-based dropout training, edge map priors, and sinusoidal positional encoding, all of which are established methods previously limited to 2D domains. We show that our framework significantly improves the quality of the resulting synthetic images and is adaptable to unseen data with fine-tuning. We also demonstrate that our resulting dataset can be employed for accuracy validation of cortical segmentation and thickness measurement.
comment: Accepted at Medical Imaging with Deep Learning (MIDL) 2023
☆ NFL Career Success as Predicted by NFL Scouting Combine
The National Football League (NFL) Scouting Combine serves as a tool to evaluate the skills of prospective players and assess their readiness to play in the NFL. The development of machine learning brings new opportunities in assessing the utility of the Scouting Combine. Using machine and statistical learning, it may be possible to predict future success of prospective athletes, as well as predict which Scouting Combine tests are the most important. Results from statistical learning research have been contradicting whether the Scouting combine is a useful metric for player success. In this study, we investigate if machine learning can be used to determine matriculation and future success in the NFL. Using Scouting Combine data, we evaluate six different algorithms' ability to predict whether a potential draft pick will play a single NFL snap (matriculation). If a player is drafted, we predict how many snaps they go on to play (success). We are able to predict matriculation with 83% accuracy; however, we are unable to predict later success. Our best performing algorithm returns large error and low explained variance (RMSE=1,210 snaps; ${R}^2$=0.17). These findings indicate that while the Scouting Combine can predict NFL matriculation, it may not be a reliable predictor of long-term player success.
comment: 16 pages, 2 figures, 1 table
☆ TrojDiff: Trojan Attacks on Diffusion Models with Diverse Targets CVPR2023
Diffusion models have achieved great success in a range of tasks, such as image synthesis and molecule design. As such successes hinge on large-scale training data collected from diverse sources, the trustworthiness of these collected data is hard to control or audit. In this work, we aim to explore the vulnerabilities of diffusion models under potential training data manipulations and try to answer: How hard is it to perform Trojan attacks on well-trained diffusion models? What are the adversarial targets that such Trojan attacks can achieve? To answer these questions, we propose an effective Trojan attack against diffusion models, TrojDiff, which optimizes the Trojan diffusion and generative processes during training. In particular, we design novel transitions during the Trojan diffusion process to diffuse adversarial targets into a biased Gaussian distribution and propose a new parameterization of the Trojan generative process that leads to an effective training objective for the attack. In addition, we consider three types of adversarial targets: the Trojaned diffusion models will always output instances belonging to a certain class from the in-domain distribution (In-D2D attack), out-of-domain distribution (Out-D2D-attack), and one specific instance (D2I attack). We evaluate TrojDiff on CIFAR-10 and CelebA datasets against both DDPM and DDIM diffusion models. We show that TrojDiff always achieves high attack performance under different adversarial targets using different types of triggers, while the performance in benign environments is preserved. The code is available at https://github.com/chenweixin107/TrojDiff.
comment: CVPR2023
☆ Fast Diffusion Sampler for Inverse Problems by Geometric Decomposition
Diffusion models have shown exceptional performance in solving inverse problems. However, one major limitation is the slow inference time. While faster diffusion samplers have been developed for unconditional sampling, there has been limited research on conditional sampling in the context of inverse problems. In this study, we propose a novel and efficient diffusion sampling strategy that employs the geometric decomposition of diffusion sampling. Specifically, we discover that the samples generated from diffusion models can be decomposed into two orthogonal components: a ``denoised" component obtained by projecting the sample onto the clean data manifold, and a ``noise" component that induces a transition to the next lower-level noisy manifold with the addition of stochastic noise. Furthermore, we prove that, under some conditions on the clean data manifold, the conjugate gradient update for imposing conditioning from the denoised signal belongs to the clean manifold, resulting in a much faster and more accurate diffusion sampling. Our method is applicable regardless of the parameterization and setting (i.e., VE, VP). Notably, we achieve state-of-the-art reconstruction quality on challenging real-world medical inverse imaging problems, including multi-coil MRI reconstruction and 3D CT reconstruction. Moreover, our proposed method achieves more than 80 times faster inference time than the previous state-of-the-art method.
comment: 21 pages
☆ Phase Aberration Correction without Reference Data: An Adaptive Mixed Loss Deep Learning Approach
Phase aberration is one of the primary sources of image quality degradation in ultrasound, which is induced by spatial variations in sound speed across the heterogeneous medium. This effect disrupts transmitted waves and prevents coherent summation of echo signals, resulting in suboptimal image quality. In real experiments, obtaining non-aberrated ground truths can be extremely challenging, if not infeasible. It hinders the performance of deep learning-based phase aberration correction techniques due to sole reliance on simulated data and the presence of domain shift between simulated and experimental data. Here, for the first time, we propose a deep learning-based method that does not require reference data to compensate for the phase aberration effect. We train a network wherein both input and target output are randomly aberrated radio frequency (RF) data. Moreover, we demonstrate that a conventional loss function such as mean square error is inadequate for training the network to achieve optimal performance. Instead, we propose an adaptive mixed loss function that employs both B-mode and RF data, resulting in more efficient convergence and enhanced performance. Source code is available at \url{http://code.sonography.ai}.
☆ Boosting Semi-Supervised Few-Shot Object Detection with SoftER Teacher
Few-shot object detection is an emerging problem aimed at detecting novel concepts from few exemplars. Existing approaches to few-shot detection assume abundant base labels to adapt to novel objects. This paper explores the task of semi-supervised few-shot detection by considering a realistic scenario which lacks abundant labels for both base and novel objects. Motivated by this unique problem, we introduce SoftER Teacher, a robust detector combining the advantages of pseudo-labeling with representation learning on region proposals. SoftER Teacher harnesses unlabeled data to jointly optimize for semi-supervised few-shot detection without explicitly relying on abundant base labels. Extensive experiments show that SoftER Teacher matches the novel class performance of a strong supervised detector using only 10% of base labels. Our work also sheds insight into a previously unknown relationship between semi-supervised and few-shot detection to suggest that a stronger semi-supervised detector leads to a more label-efficient few-shot detector. Code and models are available at https://github.com/lexisnexis-risk-open-source/ledetection
comment: Technical Report
☆ Clinical BERTScore: An Improved Measure of Automatic Speech Recognition Performance in Clinical Settings
Automatic Speech Recognition (ASR) in medical contexts has the potential to save time, cut costs, increase report accuracy, and reduce physician burnout. However, the healthcare industry has been slower to adopt this technology, in part due to the importance of avoiding medically-relevant transcription mistakes. In this work, we present the Clinical BERTScore (CBERTScore), an ASR metric that penalizes clinically-relevant mistakes more than others. We demonstrate that this metric more closely aligns with clinician preferences on medical sentences as compared to other metrics (WER, BLUE, METEOR, etc), sometimes by wide margins. We collect a benchmark of 13 clinician preferences on 149 realistic medical sentences called the Clinician Transcript Preference benchmark (CTP), demonstrate that CBERTScore more closely matches what clinicians prefer, and release the benchmark for the community to further develop clinically-aware ASR metrics.
☆ Hardware Acceleration of Neural Graphics
Rendering and inverse-rendering algorithms that drive conventional computer graphics have recently been superseded by neural representations (NR). NRs have recently been used to learn the geometric and the material properties of the scenes and use the information to synthesize photorealistic imagery, thereby promising a replacement for traditional rendering algorithms with scalable quality and predictable performance. In this work we ask the question: Does neural graphics (NG) need hardware support? We studied representative NG applications showing that, if we want to render 4k res. at 60FPS there is a gap of 1.5X-55X in the desired performance on current GPUs. For AR/VR applications, there is an even larger gap of 2-4 OOM between the desired performance and the required system power. We identify that the input encoding and the MLP kernels are the performance bottlenecks, consuming 72%,60% and 59% of application time for multi res. hashgrid, multi res. densegrid and low res. densegrid encodings, respectively. We propose a NG processing cluster, a scalable and flexible hardware architecture that directly accelerates the input encoding and MLP kernels through dedicated engines and supports a wide range of NG applications. We also accelerate the rest of the kernels by fusing them together in Vulkan, which leads to 9.94X kernel-level performance improvement compared to un-fused implementation of the pre-processing and the post-processing kernels. Our results show that, NGPC gives up to 58X end-to-end application-level performance improvement, for multi res. hashgrid encoding on average across the four NG applications, the performance benefits are 12X,20X,33X and 39X for the scaling factor of 8,16,32 and 64, respectively. Our results show that with multi res. hashgrid encoding, NGPC enables the rendering of 4k res. at 30FPS for NeRF and 8k res. at 120FPS for all our other NG applications.
☆ Provably Efficient Model-Free Algorithms for Non-stationary CMDPs
We study model-free reinforcement learning (RL) algorithms in episodic non-stationary constrained Markov Decision Processes (CMDPs), in which an agent aims to maximize the expected cumulative reward subject to a cumulative constraint on the expected utility (cost). In the non-stationary environment, reward, utility functions, and transition kernels can vary arbitrarily over time as long as the cumulative variations do not exceed certain variation budgets. We propose the first model-free, simulator-free RL algorithms with sublinear regret and zero constraint violation for non-stationary CMDPs in both tabular and linear function approximation settings with provable performance guarantees. Our results on regret bound and constraint violation for the tabular case match the corresponding best results for stationary CMDPs when the total budget is known. Additionally, we present a general framework for addressing the well-known challenges associated with analyzing non-stationary CMDPs, without requiring prior knowledge of the variation budget. We apply the approach for both tabular and linear approximation settings.
☆ Upper Bound of Real Log Canonical Threshold of Tensor Decomposition and its Application to Bayesian Inference
Tensor decomposition is now being used for data analysis, information compression, and knowledge recovery. However, the mathematical property of tensor decomposition is not yet fully clarified because it is one of singular learning machines. In this paper, we give the upper bound of its real log canonical threshold (RLCT) of the tensor decomposition by using an algebraic geometrical method and derive its Bayesian generalization error theoretically. We also give considerations about its mathematical property through numerical experiments.
☆ Explaining Model Confidence Using Counterfactuals AAAI 2023
Displaying confidence scores in human-AI interaction has been shown to help build trust between humans and AI systems. However, most existing research uses only the confidence score as a form of communication. As confidence scores are just another model output, users may want to understand why the algorithm is confident to determine whether to accept the confidence score. In this paper, we show that counterfactual explanations of confidence scores help study participants to better understand and better trust a machine learning model's prediction. We present two methods for understanding model confidence using counterfactual explanation: (1) based on counterfactual examples; and (2) based on visualisation of the counterfactual space. Both increase understanding and trust for study participants over a baseline of no explanation, but qualitative results show that they are used quite differently, leading to recommendations of when to use each one and directions of designing better explanations.
comment: AAAI 2023 Main Track. arXiv admin note: substantial text overlap with arXiv:2206.02790
☆ On the effectiveness of neural priors in modeling dynamical systems
Modelling dynamical systems is an integral component for understanding the natural world. To this end, neural networks are becoming an increasingly popular candidate owing to their ability to learn complex functions from large amounts of data. Despite this recent progress, there has not been an adequate discussion on the architectural regularization that neural networks offer when learning such systems, hindering their efficient usage. In this paper, we initiate a discussion in this direction using coordinate networks as a test bed. We interpret dynamical systems and coordinate networks from a signal processing lens, and show that simple coordinate networks with few layers can be used to solve multiple problems in modelling dynamical systems, without any explicit regularizers.
☆ Boosting Adversarial Attacks by Leveraging Decision Boundary Information
Due to the gap between a substitute model and a victim model, the gradient-based noise generated from a substitute model may have low transferability for a victim model since their gradients are different. Inspired by the fact that the decision boundaries of different models do not differ much, we conduct experiments and discover that the gradients of different models are more similar on the decision boundary than in the original position. Moreover, since the decision boundary in the vicinity of an input image is flat along most directions, we conjecture that the boundary gradients can help find an effective direction to cross the decision boundary of the victim models. Based on it, we propose a Boundary Fitting Attack to improve transferability. Specifically, we introduce a method to obtain a set of boundary points and leverage the gradient information of these points to update the adversarial examples. Notably, our method can be combined with existing gradient-based methods. Extensive experiments prove the effectiveness of our method, i.e., improving the success rate by 5.6% against normally trained CNNs and 14.9% against defense CNNs on average compared to state-of-the-art transfer-based attacks. Further we compare transformers with CNNs, the results indicate that transformers are more robust than CNNs. However, our method still outperforms existing methods when attacking transformers. Specifically, when using CNNs as substitute models, our method obtains an average attack success rate of 58.2%, which is 10.8% higher than other state-of-the-art transfer-based attacks.
☆ Tradeoff of generalization error in unsupervised learning
Finding the optimal model complexity that minimizes the generalization error (GE) is a key issue of machine learning. For the conventional supervised learning, this task typically involves the bias-variance tradeoff: lowering the bias by making the model more complex entails an increase in the variance. Meanwhile, little has been studied about whether the same tradeoff exists for unsupervised learning. In this study, we propose that unsupervised learning generally exhibits a two-component tradeoff of the GE, namely the model error and the data error -- using a more complex model reduces the model error at the cost of the data error, with the data error playing a more significant role for a smaller training dataset. This is corroborated by training the restricted Boltzmann machine to generate the configurations of the two-dimensional Ising model at a given temperature and the totally asymmetric simple exclusion process with given entry and exit rates. Our results also indicate that the optimal model tends to be more complex when the data to be learned are more complex.
comment: 15 pages, 7 figures
☆ A Unified and Efficient Coordinating Framework for Autonomous DBMS Tuning SIGMOD '23
Recently using machine learning (ML) based techniques to optimize modern database management systems has attracted intensive interest from both industry and academia. With an objective to tune a specific component of a DBMS (e.g., index selection, knobs tuning), the ML-based tuning agents have shown to be able to find better configurations than experienced database administrators. However, one critical yet challenging question remains unexplored -- how to make those ML-based tuning agents work collaboratively. Existing methods do not consider the dependencies among the multiple agents, and the model used by each agent only studies the effect of changing the configurations in a single component. To tune different components for DBMS, a coordinating mechanism is needed to make the multiple agents cognizant of each other. Also, we need to decide how to allocate the limited tuning budget among the agents to maximize the performance. Such a decision is difficult to make since the distribution of the reward for each agent is unknown and non-stationary. In this paper, we study the above question and present a unified coordinating framework to efficiently utilize existing ML-based agents. First, we propose a message propagation protocol that specifies the collaboration behaviors for agents and encapsulates the global tuning messages in each agent's model. Second, we combine Thompson Sampling, a well-studied reinforcement learning algorithm with a memory buffer so that our framework can allocate budget judiciously in a non-stationary environment. Our framework defines the interfaces adapted to a broad class of ML-based tuning agents, yet simple enough for integration with existing implementations and future extensions. We show that it can effectively utilize different ML-based agents and find better configurations with 1.4~14.1X speedups on the workload execution time compared with baselines.
comment: Accepted at 2023 International Conference on Management of Data (SIGMOD '23)
☆ Feature Unlearning for Generative Models via Implicit Feedback
We tackle the problem of feature unlearning from a pretrained image generative model. Unlike a common unlearning task where an unlearning target is a subset of the training set, we aim to unlearn a specific feature, such as hairstyle from facial images, from the pretrained generative models. As the target feature is only presented in a local region of an image, unlearning the entire image from the pretrained model may result in losing other details in the remaining region of the image. To specify which features to unlearn, we develop an implicit feedback mechanism where a user can select images containing the target feature. From the implicit feedback, we identify a latent representation corresponding to the target feature and then use the representation to unlearn the generative model. Our framework is generalizable for the two well-known families of generative models: GANs and VAEs. Through experiments on MNIST and CelebA datasets, we show that target features are successfully removed while keeping the fidelity of the original models.
☆ Fairness-enhancing deep learning for ride-hailing demand prediction
Short-term demand forecasting for on-demand ride-hailing services is one of the fundamental issues in intelligent transportation systems. However, previous travel demand forecasting research predominantly focused on improving prediction accuracy, ignoring fairness issues such as systematic underestimations of travel demand in disadvantaged neighborhoods. This study investigates how to measure, evaluate, and enhance prediction fairness between disadvantaged and privileged communities in spatial-temporal demand forecasting of ride-hailing services. A two-pronged approach is taken to reduce the demand prediction bias. First, we develop a novel deep learning model architecture, named socially aware neural network (SA-Net), to integrate the socio-demographics and ridership information for fair demand prediction through an innovative socially-aware convolution operation. Second, we propose a bias-mitigation regularization method to mitigate the mean percentage prediction error gap between different groups. The experimental results, validated on the real-world Chicago Transportation Network Company (TNC) data, show that the de-biasing SA-Net can achieve better predictive performance in both prediction accuracy and fairness. Specifically, the SA-Net improves prediction accuracy for both the disadvantaged and privileged groups compared with the state-of-the-art models. When coupled with the bias mitigation regularization method, the de-biasing SA-Net effectively bridges the mean percentage prediction error gap between the disadvantaged and privileged groups, and also protects the disadvantaged regions against systematic underestimation of TNC demand. Our proposed de-biasing method can be adopted in many existing short-term travel demand estimation models, and can be utilized for various other spatial-temporal prediction tasks such as crime incidents predictions.
☆ Explainable Semantic Medical Image Segmentation with Style
Semantic medical image segmentation using deep learning has recently achieved high accuracy, making it appealing to clinical problems such as radiation therapy. However, the lack of high-quality semantically labelled data remains a challenge leading to model brittleness to small shifts to input data. Most works require extra data for semi-supervised learning and lack the interpretability of the boundaries of the training data distribution during training, which is essential for model deployment in clinical practice. We propose a fully supervised generative framework that can achieve generalisable segmentation with only limited labelled data by simultaneously constructing an explorable manifold during training. The proposed approach creates medical image style paired with a segmentation task driven discriminator incorporating end-to-end adversarial training. The discriminator is generalised to small domain shifts as much as permissible by the training data, and the generator automatically diversifies the training samples using a manifold of input features learnt during segmentation. All the while, the discriminator guides the manifold learning by supervising the semantic content and fine-grained features separately during the image diversification. After training, visualisation of the learnt manifold from the generator is available to interpret the model limits. Experiments on a fully semantic, publicly available pelvis dataset demonstrated that our method is more generalisable to shifts than other state-of-the-art methods while being more explainable using an explorable manifold.
☆ Gaussian Max-Value Entropy Search for Multi-Agent Bayesian Optimization
We study the multi-agent Bayesian optimization (BO) problem, where multiple agents maximize a black-box function via iterative queries. We focus on Entropy Search (ES), a sample-efficient BO algorithm that selects queries to maximize the mutual information about the maximum of the black-box function. One of the main challenges of ES is that calculating the mutual information requires computationally-costly approximation techniques. For multi-agent BO problems, the computational cost of ES is exponential in the number of agents. To address this challenge, we propose the Gaussian Max-value Entropy Search, a multi-agent BO algorithm with favorable sample and computational efficiency. The key to our idea is to use a normal distribution to approximate the function maximum and calculate its mutual information accordingly. The resulting approximation allows queries to be cast as the solution of a closed-form optimization problem which, in turn, can be solved via a modified gradient ascent algorithm and scaled to a large number of agents. We demonstrate the effectiveness of Gaussian max-value Entropy Search through numerical experiments on standard test functions and real-robot experiments on the source-seeking problem. Results show that the proposed algorithm outperforms the multi-agent BO baselines in the numerical experiments and can stably seek the source with a limited number of noisy observations on real robots.
comment: 10 pages, 9 figures
☆ Human Pose Estimation from Ambiguous Pressure Recordings with Spatio-temporal Masked Transformers
Despite the impressive performance of vision-based pose estimators, they generally fail to perform well under adverse vision conditions and often don't satisfy the privacy demands of customers. As a result, researchers have begun to study tactile sensing systems as an alternative. However, these systems suffer from noisy and ambiguous recordings. To tackle this problem, we propose a novel solution for pose estimation from ambiguous pressure data. Our method comprises a spatio-temporal vision transformer with an encoder-decoder architecture. Detailed experiments on two popular public datasets reveal that our model outperforms existing solutions in the area. Moreover, we observe that increasing the number of temporal crops in the early stages of the network positively impacts the performance while pre-training the network in a self-supervised setting using a masked auto-encoder approach also further improves the results.
☆ Hierarchical Clustering with OWA-based Linkages, the Lance-Williams Formula, and Dendrogram Inversions
Agglomerative hierarchical clustering based on Ordered Weighted Averaging (OWA) operators not only generalises the single, complete, and average linkages, but also includes intercluster distances based on a few nearest or farthest neighbours, trimmed and winsorised means of pairwise point similarities, amongst many others. We explore the relationships between the famous Lance-Williams update formula and the extended OWA-based linkages with weights generated via infinite coefficient sequences. Furthermore, we provide some conditions for the weight generators to guarantee the resulting dendrograms to be free from unaesthetic inversions.
☆ A dual basis approach to multidimensional scaling: spectral analysis and graph regularity
Classical multidimensional scaling (CMDS) is a technique that aims to embed a set of objects in a Euclidean space given their pairwise Euclidean distance matrix. The main part of CMDS is based on double centering a squared distance matrix and employing a truncated eigendecomposition to recover the point coordinates. A central result in CMDS connects the squared Euclidean matrix to a Gram matrix derived from the set of points. In this paper, we study a dual basis approach to classical multidimensional scaling. We give an explicit formula for the dual basis and fully characterize the spectrum of an essential matrix in the dual basis framework. We make connections to a related problem in metric nearness.
comment: 9 pages
☆ Clustering with minimum spanning trees: How good can it be?
Minimum spanning trees (MSTs) provide a convenient representation of datasets in numerous pattern recognition activities. Moreover, they are relatively fast to compute. In this paper, we quantify the extent to which they can be meaningful in data clustering tasks. By identifying the upper bounds for the agreement between the best (oracle) algorithm and the expert labels from a large battery of benchmark data, we discover that MST methods can overall be very competitive. Next, instead of proposing yet another algorithm that performs well on a limited set of examples, we review, study, extend, and generalise existing, the state-of-the-art MST-based partitioning schemes, which leads to a few new and interesting approaches. It turns out that the Genie method and the information-theoretic approaches often outperform the non-MST algorithms such as k-means, Gaussian mixtures, spectral clustering, BIRCH, and classical hierarchical agglomerative procedures.
☆ Improving Weakly Supervised Sound Event Detection with Causal Intervention ICASSP2023
Existing weakly supervised sound event detection (WSSED) work has not explored both types of co-occurrences simultaneously, i.e., some sound events often co-occur, and their occurrences are usually accompanied by specific background sounds, so they would be inevitably entangled, causing misclassification and biased localization results with only clip-level supervision. To tackle this issue, we first establish a structural causal model (SCM) to reveal that the context is the main cause of co-occurrence confounders that mislead the model to learn spurious correlations between frames and clip-level labels. Based on the causal analysis, we propose a causal intervention (CI) method for WSSED to remove the negative impact of co-occurrence confounders by iteratively accumulating every possible context of each class and then re-projecting the contexts to the frame-level features for making the event boundary clearer. Experiments show that our method effectively improves the performance on multiple datasets and can generalize to various baseline models.
comment: Accepted by ICASSP2023
☆ Towards better traffic volume estimation: Tackling both underdetermined and non-equilibrium problems via a correlation adaptive graph convolution network
Traffic volume is an indispensable ingredient to provide fine-grained information for traffic management and control. However, due to limited deployment of traffic sensors, obtaining full-scale volume information is far from easy. Existing works on this topic primarily focus on improving the overall estimation accuracy of a particular method and ignore the underlying challenges of volume estimation, thereby having inferior performances on some critical tasks. This paper studies two key problems with regard to traffic volume estimation: (1) underdetermined traffic flows caused by undetected movements, and (2) non-equilibrium traffic flows arise from congestion propagation. Here we demonstrate a graph-based deep learning method that can offer a data-driven, model-free and correlation adaptive approach to tackle the above issues and perform accurate network-wide traffic volume estimation. Particularly, in order to quantify the dynamic and nonlinear relationships between traffic speed and volume for the estimation of underdetermined flows, a speed patternadaptive adjacent matrix based on graph attention is developed and integrated into the graph convolution process, to capture non-local correlations between sensors. To measure the impacts of non-equilibrium flows, a temporal masked and clipped attention combined with a gated temporal convolution layer is customized to capture time-asynchronous correlations between upstream and downstream sensors. We then evaluate our model on a real-world highway traffic volume dataset and compare it with several benchmark models. It is demonstrated that the proposed model achieves high estimation accuracy even under 20% sensor coverage rate and outperforms other baselines significantly, especially on underdetermined and non-equilibrium flow locations. Furthermore, comprehensive quantitative model analysis are also carried out to justify the model designs.
☆ EHRDiff: Exploring Realistic EHR Synthesis with Diffusion Models
Electronic health records (EHR) contain vast biomedical knowledge and are rich resources for developing precise medicine systems. However, due to privacy concerns, there are limited high-quality EHR data accessible to researchers hence hindering the advancement of methodologies. Recent research has explored using generative modelling methods to synthesize realistic EHR data, and most proposed methods are based on the generative adversarial network (GAN) and its variants for EHR synthesis. Although GAN-style methods achieved state-of-the-art performance in generating high-quality EHR data, such methods are hard to train and prone to mode collapse. Diffusion models are recently proposed generative modelling methods and set cutting-edge performance in image generation. The performance of diffusion models in realistic EHR synthesis is rarely explored. In this work, we explore whether the superior performance of diffusion models can translate to the domain of EHR synthesis and propose a novel EHR synthesis method named EHRDiff. Through comprehensive experiments, EHRDiff achieves new state-of-the-art performance for the quality of synthetic EHR data and can better protect private information in real training EHRs in the meanwhile.
comment: Working in progress
☆ GATOR: Graph-Aware Transformer with Motion-Disentangled Regression for Human Mesh Recovery from a 2D Pose ICASSP 2023
3D human mesh recovery from a 2D pose plays an important role in various applications. However, it is hard for existing methods to simultaneously capture the multiple relations during the evolution from skeleton to mesh, including joint-joint, joint-vertex and vertex-vertex relations, which often leads to implausible results. To address this issue, we propose a novel solution, called GATOR, that contains an encoder of Graph-Aware Transformer (GAT) and a decoder with Motion-Disentangled Regression (MDR) to explore these multiple relations. Specifically, GAT combines a GCN and a graph-aware self-attention in parallel to capture physical and hidden joint-joint relations. Furthermore, MDR models joint-vertex and vertex-vertex interactions to explore joint and vertex relations. Based on the clustering characteristics of vertex offset fields, MDR regresses the vertices by composing the predicted base motions. Extensive experiments show that GATOR achieves state-of-the-art performance on two challenging benchmarks.
comment: Accepted by ICASSP 2023
☆ Pacos: Modeling Users' Interpretable and Context-Dependent Choices in Preference Reversals
Choice problems refer to selecting the best choices from several items, and learning users' preferences in choice problems is of great significance in understanding the decision making mechanisms and providing personalized services. Existing works typically assume that people evaluate items independently. In practice, however, users' preferences depend on the market in which items are placed, which is known as context effects; and the order of users' preferences for two items may even be reversed, which is referred to preference reversals. In this work, we identify three factors contributing to context effects: users' adaptive weights, the inter-item comparison, and display positions. We propose a context-dependent preference model named Pacos as a unified framework for addressing three factors simultaneously, and consider two design methods including an additive method with high interpretability and an ANN-based method with high accuracy. We study the conditions for preference reversals to occur and provide an theoretical proof of the effectiveness of Pacos in addressing preference reversals. Experimental results show that the proposed method has better performance than prior works in predicting users' choices, and has great interpretability to help understand the cause of preference reversals.
comment: 29 pages, 12 figures
☆ Efficient Real Time Recurrent Learning through combined activity and parameter sparsity ICLR 2023
Backpropagation through time (BPTT) is the standard algorithm for training recurrent neural networks (RNNs), which requires separate simulation phases for the forward and backward passes for inference and learning, respectively. Moreover, BPTT requires storing the complete history of network states between phases, with memory consumption growing proportional to the input sequence length. This makes BPTT unsuited for online learning and presents a challenge for implementation on low-resource real-time systems. Real-Time Recurrent Learning (RTRL) allows online learning, and the growth of required memory is independent of sequence length. However, RTRL suffers from exceptionally high computational costs that grow proportional to the fourth power of the state size, making RTRL computationally intractable for all but the smallest of networks. In this work, we show that recurrent networks exhibiting high activity sparsity can reduce the computational cost of RTRL. Moreover, combining activity and parameter sparsity can lead to significant enough savings in computational and memory costs to make RTRL practical. Unlike previous work, this improvement in the efficiency of RTRL can be achieved without using any approximations for the learning process.
comment: Published as a workshop paper at ICLR 2023 Workshop on Sparsity in Neural Networks
Self-Supervised One-Shot Learning for Automatic Segmentation of StyleGAN Images
We propose in this paper a framework for automatic one-shot segmentation of synthetic images generated using StyleGANs. As to the need for `one-shot segmentation', we want the network to carry out a semantic segmentation of the images on the fly, that is, as they are being produced at inference time. The implementation of our framework is based on the observation that the multi-scale hidden features produced by a GAN during image synthesis hold useful semantic information that can be utilized for automatic segmentation. Using these features, our proposed framework learns to segment synthetic images using a novel self-supervised, contrastive clustering algorithm that projects the hidden features in the generator onto a compact feature space for per-pixel classification. This contrastive learner uses a swapped prediction loss for image segmentation that is computed using pixel-wise cluster assignments for the image and its transformed variants. Using the hidden features from an already pre-trained GAN for clustering, this leads to a much faster learning of the pixel-wise feature vectors for one-shot segmentation. We have tested our implementation on a number of standard benchmarks (CelebA, LSUN, PASCAL-Part) for object and part segmentation. The results of our experiments yield a segmentation performance that not only outperforms the semi-supervised baseline methods with an average wIoU margin of 1.02 % but also improves the inference speeds by a peak factor of 4.5. Finally, we also show the results of using the proposed framework in the implementation of BagGAN, a GAN-based framework for the production of annotated synthetic baggage X-ray scans for threat detection. This one-shot learning framework was trained and tested on the PIDRay baggage screening benchmark for 5 different threat categories to yield a segmentation performance which stands close to its baseline segmenter.
☆ Fusarium head blight detection, spikelet estimation, and severity assessment in wheat using 3D convolutional neural networks
Fusarium head blight (FHB) is one of the most significant diseases affecting wheat and other small grain cereals worldwide. The development of resistant varieties requires the laborious task of field and greenhouse phenotyping. The applications considered in this work are the automated detection of FHB disease symptoms expressed on a wheat plant, the automated estimation of the total number of spikelets and the total number of infected spikelets on a wheat head, and the automated assessment of the FHB severity in infected wheat. The data used to generate the results are 3-dimensional (3D) multispectral point clouds (PC), which are 3D collections of points - each associated with a red, green, blue (RGB), and near-infrared (NIR) measurement. Over 300 wheat plant images were collected using a multispectral 3D scanner, and the labelled UW-MRDC 3D wheat dataset was created. The data was used to develop novel and efficient 3D convolutional neural network (CNN) models for FHB detection, which achieved 100% accuracy. The influence of the multispectral information on performance was evaluated, and our results showed the dominance of the RGB channels over both the NIR and the NIR plus RGB channels combined. Furthermore, novel and efficient 3D CNNs were created to estimate the total number of spikelets and the total number of infected spikelets on a wheat head, and our best models achieved mean absolute errors (MAE) of 1.13 and 1.56, respectively. Moreover, 3D CNN models for FHB severity estimation were created, and our best model achieved 8.6 MAE. A linear regression analysis between the visual FHB severity assessment and the FHB severity predicted by our 3D CNN was performed, and the results showed a significant correlation between the two variables with a 0.0001 P-value and 0.94 R-squared.
☆ Monitoring Efficiency of IoT Wireless Charging
Crowdsourcing wireless energy is a novel and convenient solution to charge nearby IoT devices. Several applications have been proposed to enable peer-to-peer wireless energy charging. However, none of them considered the energy efficiency of the wireless transfer of energy. In this paper, we propose an energy estimation framework that predicts the actual received energy. Our framework uses two machine learning algorithms, namely XGBoost and Neural Network, to estimate the received energy. The result shows that the Neural Network model is better than XGBoost at predicting the received energy. We train and evaluate our models by collecting a real wireless energy dataset.
comment: 3 pages, 4 figures. This is an accepted demo paper and it will appear in The 21st International Conference on Pervasive Computing and Communications (PerCom 2023)
☆ On the Unlikelihood of D-Separation
Causal discovery aims to recover a causal graph from data generated by it; constraint based methods do so by searching for a d-separating conditioning set of nodes in the graph via an oracle. In this paper, we provide analytic evidence that on large graphs, d-separation is a rare phenomenon, even when guaranteed to exist, unless the graph is extremely sparse. We then provide an analytic average case analysis of the PC Algorithm for causal discovery, as well as a variant of the SGS Algorithm we call UniformSGS. We consider a set $V=\{v_1,\ldots,v_n\}$ of nodes, and generate a random DAG $G=(V,E)$ where $(v_a, v_b) \in E$ with i.i.d. probability $p_1$ if $a b$. We provide upper bounds on the probability that a subset of $V-\{x,y\}$ d-separates $x$ and $y$, conditional on $x$ and $y$ being d-separable; our upper bounds decay exponentially fast to $0$ as $|V| \rightarrow \infty$. For the PC Algorithm, while it is known that its worst-case guarantees fail on non-sparse graphs, we show that the same is true for the average case, and that the sparsity requirement is quite demanding: for good performance, the density must go to $0$ as $|V| \rightarrow \infty$ even in the average case. For UniformSGS, while it is known that the running time is exponential for existing edges, we show that in the average case, that is the expected running time for most non-existing edges as well.
♻ ☆ An analytic theory for the dynamics of wide quantum neural networks
Parameterized quantum circuits can be used as quantum neural networks and have the potential to outperform their classical counterparts when trained for addressing learning problems. To date, much of the results on their performance on practical problems are heuristic in nature. In particular, the convergence rate for the training of quantum neural networks is not fully understood. Here, we analyze the dynamics of gradient descent for the training error of a class of variational quantum machine learning models. We define wide quantum neural networks as parameterized quantum circuits in the limit of a large number of qubits and variational parameters. We then find a simple analytic formula that captures the average behavior of their loss function and discuss the consequences of our findings. For example, for random quantum circuits, we predict and characterize an exponential decay of the residual training error as a function of the parameters of the system. We finally validate our analytic results with numerical experiments.
comment: 37 pages, many figures. v2: adding learning supervised perspectives and new results, close to published version
♻ ☆ DDPNAS: Efficient Neural Architecture Search via Dynamic Distribution Pruning
Neural Architecture Search (NAS) has demonstrated state-of-the-art performance on various computer vision tasks. Despite the superior performance achieved, the efficiency and generality of existing methods are highly valued due to their high computational complexity and low generality. In this paper, we propose an efficient and unified NAS framework termed DDPNAS via dynamic distribution pruning, facilitating a theoretical bound on accuracy and efficiency. In particular, we first sample architectures from a joint categorical distribution. Then the search space is dynamically pruned and its distribution is updated every few epochs. With the proposed efficient network generation method, we directly obtain the optimal neural architectures on given constraints, which is practical for on-device models across diverse search spaces and constraints. The architectures searched by our method achieve remarkable top-1 accuracies, 97.56 and 77.2 on CIFAR-10 and ImageNet (mobile settings), respectively, with the fastest search process, i.e., only 1.8 GPU hours on a Tesla V100. Codes for searching and network generation are available at: https://openi.pcl.ac.cn/PCL AutoML/XNAS.
comment: A update version of this work. 19 pages
♻ ☆ DEJA VU: Continual Model Generalization For Unseen Domains ICLR 2023
In real-world applications, deep learning models often run in non-stationary environments where the target data distribution continually shifts over time. There have been numerous domain adaptation (DA) methods in both online and offline modes to improve cross-domain adaptation ability. However, these DA methods typically only provide good performance after a long period of adaptation, and perform poorly on new domains before and during adaptation - in what we call the "Unfamiliar Period", especially when domain shifts happen suddenly and significantly. On the other hand, domain generalization (DG) methods have been proposed to improve the model generalization ability on unadapted domains. However, existing DG works are ineffective for continually changing domains due to severe catastrophic forgetting of learned knowledge. To overcome these limitations of DA and DG in handling the Unfamiliar Period during continual domain shift, we propose RaTP, a framework that focuses on improving models' target domain generalization (TDG) capability, while also achieving effective target domain adaptation (TDA) capability right after training on certain domains and forgetting alleviation (FA) capability on past domains. RaTP includes a training-free data augmentation module to prepare data for TDG, a novel pseudo-labeling mechanism to provide reliable supervision for TDA, and a prototype contrastive alignment algorithm to align different domains for achieving TDG, TDA and FA. Extensive experiments on Digits, PACS, and DomainNet demonstrate that RaTP significantly outperforms state-of-the-art works from Continual DA, Source-Free DA, Test-Time/Online DA, Single DG, Multiple DG and Unified DA&DG in TDG, and achieves comparable TDA and FA capabilities.
comment: Published as a conference paper at ICLR 2023
♻ ☆ Multidimensional Interactive Fixed-Effects
This paper studies a linear and additively separable model for multidimensional panel data of three or more dimensions with unobserved interactive fixed effects. Two approaches are considered to account for these unobserved interactive fixed-effects when estimating coefficients on the observed covariates. First, the model is embedded within the standard two-dimensional panel framework and restrictions are derived under which the factor structure methods in Bai (2009) lead to consistent estimation of model parameters, but at potentially slow rates of convergence. The second approach utilises popular machine learning techniques to develop group fixed-effects and kernel weighted fixed-effects that are more robust to the multidimensional nature of the problem and can achieve the parametric rate of consistency under certain conditions. Theoretical results and simulations show the benefit of standard two-dimensional panel methods when the structure of the interactive fixed-effect term is known, but also highlight how the group fixed-effects and kernel methods perform well without knowledge of this structure. The methods are implemented to estimate the demand elasticity for beer under a handful of models for demand.
♻ ☆ Deep Clustering Survival Machines with Interpretable Expert Distributions
Conventional survival analysis methods are typically ineffective to characterize heterogeneity in the population while such information can be used to assist predictive modeling. In this study, we propose a hybrid survival analysis method, referred to as deep clustering survival machines, that combines the discriminative and generative mechanisms. Similar to the mixture models, we assume that the timing information of survival data is generatively described by a mixture of certain numbers of parametric distributions, i.e., expert distributions. We learn weights of the expert distributions for individual instances according to their features discriminatively such that each instance's survival information can be characterized by a weighted combination of the learned constant expert distributions. This method also facilitates interpretable subgrouping/clustering of all instances according to their associated expert distributions. Extensive experiments on both real and synthetic datasets have demonstrated that the method is capable of obtaining promising clustering results and competitive time-to-event predicting performance.
♻ ☆ Accelerating ODE-Based Neural Networks on Low-Cost FPGAs
ODENet is a deep neural network architecture in which a stacking structure of ResNet is implemented with an ordinary differential equation (ODE) solver. It can reduce the number of parameters and strike a balance between accuracy and performance by selecting a proper solver. It is also possible to improve the accuracy while keeping the same number of parameters on resource-limited edge devices. In this paper, using Euler method as an ODE solver, a part of ODENet is implemented as a dedicated logic on a low-cost FPGA (Field-Programmable Gate Array) board, such as PYNQ-Z2 board. As ODENet variants, reduced ODENets (rODENets) each of which heavily uses a part of ODENet layers and reduces/eliminates some layers differently are proposed and analyzed for low-cost FPGA implementation. They are evaluated in terms of parameter size, accuracy, execution time, and resource utilization on the FPGA. The results show that an overall execution time of an rODENet variant is improved by up to 2.66 times compared to a pure software execution while keeping a comparable accuracy to the original ODENet.
comment: RAW'21
♻ ☆ Accelerating Distributed Deep Reinforcement Learning by In-Network Experience Sampling
A computing cluster that interconnects multiple compute nodes is used to accelerate distributed reinforcement learning based on DQN (Deep Q-Network). In distributed reinforcement learning, Actor nodes acquire experiences by interacting with a given environment and a Learner node optimizes their DQN model. Since data transfer between Actor and Learner nodes increases depending on the number of Actor nodes and their experience size, communication overhead between them is one of major performance bottlenecks. In this paper, their communication is accelerated by DPDK-based network optimizations, and DPDK-based low-latency experience replay memory server is deployed between Actor and Learner nodes interconnected with a 40GbE (40Gbit Ethernet) network. Evaluation results show that, as a network optimization technique, kernel bypassing by DPDK reduces network access latencies to a shared memory server by 32.7% to 58.9%. As another network optimization technique, an in-network experience replay memory server between Actor and Learner nodes reduces access latencies to the experience replay memory by 11.7% to 28.1% and communication latencies for prioritized experience sampling by 21.9% to 29.1%.
comment: PDP'22
♻ ☆ APTx: better activation function than MISH, SWISH, and ReLU's variants used in deep learning
Activation Functions introduce non-linearity in the deep neural networks. This nonlinearity helps the neural networks learn faster and efficiently from the dataset. In deep learning, many activation functions are developed and used based on the type of problem statement. ReLU's variants, SWISH, and MISH are goto activation functions. MISH function is considered having similar or even better performance than SWISH, and much better than ReLU. In this paper, we propose an activation function named APTx which behaves similar to MISH, but requires lesser mathematical operations to compute. The lesser computational requirements of APTx does speed up the model training, and thus also reduces the hardware requirement for the deep learning model.
comment: 8 pages, 6 figures
♻ ☆ Communication Size Reduction of Federated Learning using Neural ODE Models
Federated learning is a machine learning approach in which data is not aggregated on a server, but is trained at clients locally, in consideration of security and privacy. ResNet is a classic but representative neural network that succeeds in deepening the neural network by learning a residual function that adds the inputs and outputs together. In federated learning, communication is performed between the server and clients to exchange weight parameters. Since ResNet has deep layers and a large number of parameters, the communication size becomes large. In this paper, we use Neural ODE as a lightweight model of ResNet to reduce communication size in federated learning. In addition, we newly introduce a flexible federated learning using Neural ODE models with different number of iterations, which correspond to ResNet models with different depths. Evaluation results using CIFAR-10 dataset show that the use of Neural ODE reduces communication size by up to 92.4% compared to ResNet. We also show that the proposed flexible federated learning can merge models with different iteration counts or depths.
♻ ☆ Conjugate Natural Selection: Fisher-Rao Natural Gradient Descent Optimally Approximates Evolutionary Dynamics and Continuous Bayesian Inference
Rather than refining individual candidate solutions for a general non-convex optimization problem, by analogy to evolution, we consider minimizing the average loss for a parametric distribution over hypotheses. In this setting, we prove that Fisher-Rao natural gradient descent (FR-NGD) optimally approximates the continuous-time replicator equation (an essential model of evolutionary dynamics) by minimizing the mean-squared error for the relative fitness of competing hypotheses. We term this finding "conjugate natural selection" and demonstrate its utility by numerically solving an example non-convex optimization problem over a continuous strategy space. Next, by developing known connections between discrete-time replicator dynamics and Bayes's rule, we show that when absolute fitness corresponds to the negative KL-divergence of a hypothesis's predictions from actual observations, FR-NGD provides the optimal approximation of continuous Bayesian inference. We use this result to demonstrate a novel method for estimating the parameters of stochastic processes.
comment: 13 pages, 3 figures
♻ ☆ Large Language Models Are Human-Level Prompt Engineers
By conditioning on natural language instructions, large language models (LLMs) have displayed impressive capabilities as general-purpose computers. However, task performance depends significantly on the quality of the prompt used to steer the model, and most effective prompts have been handcrafted by humans. Inspired by classical program synthesis and the human approach to prompt engineering, we propose Automatic Prompt Engineer (APE) for automatic instruction generation and selection. In our method, we treat the instruction as the "program," optimized by searching over a pool of instruction candidates proposed by an LLM in order to maximize a chosen score function. To evaluate the quality of the selected instruction, we evaluate the zero-shot performance of another LLM following the selected instruction. Experiments on 24 NLP tasks show that our automatically generated instructions outperform the prior LLM baseline by a large margin and achieve better or comparable performance to the instructions generated by human annotators on 19/24 tasks. We conduct extensive qualitative and quantitative analyses to explore the performance of APE. We show that APE-engineered prompts can be applied to steer models toward truthfulness and/or informativeness, as well as to improve few-shot learning performance by simply prepending them to standard in-context learning prompts. Please check out our webpage at https://sites.google.com/view/automatic-prompt-engineer.
♻ ☆ Reusing Combinatorial Structure: Faster Iterative Projections over Submodular Base Polytopes
Optimization algorithms such as projected Newton's method, FISTA, mirror descent, and its variants enjoy near-optimal regret bounds and convergence rates, but suffer from a computational bottleneck of computing ``projections'' in potentially each iteration (e.g., $O(T^{1/2})$ regret of online mirror descent). On the other hand, conditional gradient variants solve a linear optimization in each iteration, but result in suboptimal rates (e.g., $O(T^{3/4})$ regret of online Frank-Wolfe). Motivated by this trade-off in runtime v/s convergence rates, we consider iterative projections of close-by points over widely-prevalent submodular base polytopes $B(f)$. We first give necessary and sufficient conditions for when two close points project to the same face of a polytope, and then show that points far away from the polytope project onto its vertices with high probability. We next use this theory and develop a toolkit to speed up the computation of iterative projections over submodular polytopes using both discrete and continuous perspectives. We subsequently adapt the away-step Frank-Wolfe algorithm to use this information and enable early termination. For the special case of cardinality-based submodular polytopes, we improve the runtime of computing certain Bregman projections by a factor of $\Omega(n/\log(n))$. Our theoretical results show orders of magnitude reduction in runtime in preliminary computational experiments.
♻ ☆ Locally Regularized Neural Differential Equations: Some Black Boxes Were Meant to Remain Closed!
Implicit layer deep learning techniques, like Neural Differential Equations, have become an important modeling framework due to their ability to adapt to new problems automatically. Training a neural differential equation is effectively a search over a space of plausible dynamical systems. However, controlling the computational cost for these models is difficult since it relies on the number of steps the adaptive solver takes. Most prior works have used higher-order methods to reduce prediction timings while greatly increasing training time or reducing both training and prediction timings by relying on specific training algorithms, which are harder to use as a drop-in replacement due to strict requirements on automatic differentiation. In this manuscript, we use internal cost heuristics of adaptive differential equation solvers at stochastic time points to guide the training toward learning a dynamical system that is easier to integrate. We "close the black-box" and allow the use of our method with any adjoint technique for gradient calculations of the differential equation solution. We perform experimental studies to compare our method to global regularization to show that we attain similar performance numbers without compromising the flexibility of implementation on ordinary differential equations (ODEs) and stochastic differential equations (SDEs). We develop two sampling strategies to trade off between performance and training time. Our method reduces the number of function evaluations to 0.556-0.733x and accelerates predictions by 1.3-2x.
♻ ☆ Metrizing Fairness
We study supervised learning problems for predicting properties of individuals who belong to one of two demographic groups, and we seek predictors that are fair according to statistical parity. This means that the distributions of the predictions within the two groups should be close with respect to the Kolmogorov distance, and fairness is achieved by penalizing the dissimilarity of these two distributions in the objective function of the learning problem. In this paper, we showcase conceptual and computational benefits of measuring unfairness with integral probability metrics (IPMs) other than the Kolmogorov distance. Conceptually, we show that the generator of any IPM can be interpreted as a family of utility functions and that unfairness with respect to this IPM arises if individuals in the two demographic groups have diverging expected utilities. We also prove that the unfairness-regularized prediction loss admits unbiased gradient estimators if unfairness is measured by the squared $\mathcal L^2$-distance or by a squared maximum mean discrepancy. In this case, the fair learning problem is susceptible to efficient stochastic gradient descent (SGD) algorithms. Numerical experiments on real data show that these SGD algorithms outperform state-of-the-art methods for fair learning in that they achieve superior accuracy-unfairness trade-offs -- sometimes orders of magnitude faster. Finally, we identify conditions under which statistical parity can improve prediction accuracy.
POLICE: Provably Optimal Linear Constraint Enforcement for Deep Neural Networks
Deep Neural Networks (DNNs) outshine alternative function approximators in many settings thanks to their modularity in composing any desired differentiable operator. The formed parametrized functional is then tuned to solve a task at hand from simple gradient descent. This modularity comes at the cost of making strict enforcement of constraints on DNNs, e.g. from a priori knowledge of the task, or from desired physical properties, an open challenge. In this paper we propose the first provable affine constraint enforcement method for DNNs that only requires minimal changes into a given DNN's forward-pass, that is computationally friendly, and that leaves the optimization of the DNN's parameter to be unconstrained, i.e. standard gradient-based method can be employed. Our method does not require any sampling and provably ensures that the DNN fulfills the affine constraint on a given input space's region at any point during training, and testing. We coin this method POLICE, standing for Provably Optimal LInear Constraint Enforcement. Github: https://github.com/RandallBalestriero/POLICE
♻ ☆ Uncertainty Estimates of Predictions via a General Bias-Variance Decomposition AISTATS 2023
Reliably estimating the uncertainty of a prediction throughout the model lifecycle is crucial in many safety-critical applications. The most common way to measure this uncertainty is via the predicted confidence. While this tends to work well for in-domain samples, these estimates are unreliable under domain drift and restricted to classification. Alternatively, proper scores can be used for most predictive tasks but a bias-variance decomposition for model uncertainty does not exist in the current literature. In this work we introduce a general bias-variance decomposition for proper scores, giving rise to the Bregman Information as the variance term. We discover how exponential families and the classification log-likelihood are special cases and provide novel formulations. Surprisingly, we can express the classification case purely in the logit space. We showcase the practical relevance of this decomposition on several downstream tasks, including model ensembles and confidence regions. Further, we demonstrate how different approximations of the instance-level Bregman Information allow reliable out-of-distribution detection for all degrees of domain drift.
comment: Accepted at AISTATS 2023
♻ ☆ Model-based Causal Bayesian Optimization ICLR 2023
How should we intervene on an unknown structural equation model to maximize a downstream variable of interest? This setting, also known as causal Bayesian optimization (CBO), has important applications in medicine, ecology, and manufacturing. Standard Bayesian optimization algorithms fail to effectively leverage the underlying causal structure. Existing CBO approaches assume noiseless measurements and do not come with guarantees. We propose the model-based causal Bayesian optimization algorithm (MCBO) that learns a full system model instead of only modeling intervention-reward pairs. MCBO propagates epistemic uncertainty about the causal mechanisms through the graph and trades off exploration and exploitation via the optimism principle. We bound its cumulative regret, and obtain the first non-asymptotic bounds for CBO. Unlike in standard Bayesian optimization, our acquisition function cannot be evaluated in closed form, so we show how the reparameterization trick can be used to apply gradient-based optimizers. The resulting practical implementation of MCBO compares favorably with state-of-the-art approaches empirically.
comment: 24 pages, 8 figures, accepted at ICLR 2023
♻ ☆ Maximal Objectives in the Multi-armed Bandit with Applications
In several applications of the stochastic multi-armed bandit problem, the traditional objective of maximizing the expected total reward can be inappropriate. In this paper, motivated by certain operational concerns in online platforms, we consider a new objective in the classical setup. Given $K$ arms, instead of maximizing the expected total reward from $T$ pulls (the traditional "sum" objective), we consider the vector of total rewards earned from each of the $K$ arms at the end of $T$ pulls and aim to maximize the expected highest total reward across arms (the "max" objective). For this objective, we show that any policy must incur an instance-dependent asymptotic regret of $\Omega(\log T)$ (with a higher instance-dependent constant compared to the traditional objective) and a worst-case regret of $\Omega(K^{1/3}T^{2/3})$. We then design an adaptive explore-then-commit policy featuring exploration based on appropriately tuned confidence bounds on the mean reward and an adaptive stopping criterion, which adapts to the problem difficulty and achieves these bounds (up to logarithmic factors). We then generalize our algorithmic insights to the problem of maximizing the expected value of the average total reward of the top $m$ arms with the highest total rewards. Our numerical experiments demonstrate the efficacy of our policies compared to several natural alternatives in practical parameter regimes. We discuss applications of these new objectives to the problem of grooming an adequate supply of value-providing market participants (workers/sellers/service providers) in online platforms.
comment: 92 pages, 10 figures, 6 tables
♻ ☆ Approximate Regions of Attraction in Learning with Decision-Dependent Distributions
As data-driven methods are deployed in real-world settings, the processes that generate the observed data will often react to the decisions of the learner. For example, a data source may have some incentive for the algorithm to provide a particular label (e.g. approve a bank loan), and manipulate their features accordingly. Work in strategic classification and decision-dependent distributions seeks to characterize the closed-loop behavior of deploying learning algorithms by explicitly considering the effect of the classifier on the underlying data distribution. More recently, works in performative prediction seek to classify the closed-loop behavior by considering general properties of the mapping from classifier to data distribution, rather than an explicit form. Building on this notion, we analyze repeated risk minimization as the perturbed trajectories of the gradient flows of performative risk minimization. We consider the case where there may be multiple local minimizers of performative risk, motivated by situations where the initial conditions may have significant impact on the long-term behavior of the system. We provide sufficient conditions to characterize the region of attraction for the various equilibria in this settings. Additionally, we introduce the notion of performative alignment, which provides a geometric condition on the convergence of repeated risk minimization to performative risk minimizers.
♻ ☆ A Contrastive Approach to Online Change Point Detection AISTATS 2023
We suggest a novel procedure for online change point detection. Our approach expands an idea of maximizing a discrepancy measure between points from pre-change and post-change distributions. This leads to a flexible procedure suitable for both parametric and nonparametric scenarios. We prove non-asymptotic bounds on the average running length of the procedure and its expected detection delay. The efficiency of the algorithm is illustrated with numerical experiments on synthetic and real-world data sets.
comment: Accepted for presentation at AISTATS 2023; 28 pages
♻ ☆ DORA: Exploring outlier representations in Deep Neural Networks
Deep Neural Networks (DNNs) draw their power from the representations they learn. However, while being incredibly effective in learning complex abstractions, they are susceptible to learning malicious concepts, due to the spurious correlations inherent in the training data. So far, existing methods for uncovering such artifactual behavior in trained models focus on finding artifacts in the input data, which requires both availability of a data set and human supervision. In this paper, we introduce DORA (Data-agnOstic Representation Analysis): the first data-agnostic framework for the analysis of the representation space of DNNs. We propose a novel distance measure between representations that utilizes self-explaining capabilities within the network itself without access to any data and quantitatively validate its alignment with human-defined semantic distances. We further demonstrate that this metric could be utilized for the detection of anomalous representations, which may bear a risk of learning unintended spurious concepts deviating from the desired decision-making policy. Finally, we demonstrate the practical utility of DORA by analyzing and identifying artifactual representations in widely popular Computer Vision models.
comment: 15 pages, 11 figures
♻ ☆ Best of Many Worlds Guarantees for Online Learning with Knapsacks
We study online learning problems in which a decision maker wants to maximize their expected reward without violating a finite set of $m$ resource constraints. By casting the learning process over a suitably defined space of strategy mixtures, we recover strong duality on a Lagrangian relaxation of the underlying optimization problem, even for general settings with non-convex reward and resource-consumption functions. Then, we provide the first best-of-many-worlds type framework for this setting, with no-regret guarantees under stochastic, adversarial, and non-stationary inputs. Our framework yields the same regret guarantees of prior work in the stochastic case. On the other hand, when budgets grow at least linearly in the time horizon, it allows us to provide a constant competitive ratio in the adversarial case, which improves over the best known upper bound bound of $O(\log m \log T)$. Moreover, our framework allows the decision maker to handle non-convex reward and cost functions. We provide two game-theoretic applications of our framework to give further evidence of its flexibility. In doing so, we show that it can be employed to implement budget-pacing mechanisms in repeated first-price auctions.
♻ ☆ Machine Learning-powered Course Allocation
We introduce a machine learning-powered course allocation mechanism. Concretely, we extend the state-of-the-art Course Match mechanism with a machine learning-based preference elicitation module. In an iterative, asynchronous manner, this module generates pairwise comparison queries that are tailored to each individual student. Regarding incentives, our machine learning-powered course match (MLCM) mechanism retains the attractive strategyproofness in the large property of Course Match. Regarding welfare, we perform computational experiments using a simulator that was fitted to real-world data. Our results show that, compared to Course Match, MLCM increases average student utility by 4%-9% and minimum student utility by 10%-21%, even with only ten comparison queries. Finally, we highlight the practicability of MLCM and the ease of piloting it for universities currently using Course Match.
♻ ☆ GradMA: A Gradient-Memory-based Accelerated Federated Learning with Alleviated Catastrophic Forgetting
Federated Learning (FL) has emerged as a de facto machine learning area and received rapid increasing research interests from the community. However, catastrophic forgetting caused by data heterogeneity and partial participation poses distinctive challenges for FL, which are detrimental to the performance. To tackle the problems, we propose a new FL approach (namely GradMA), which takes inspiration from continual learning to simultaneously correct the server-side and worker-side update directions as well as take full advantage of server's rich computing and memory resources. Furthermore, we elaborate a memory reduction strategy to enable GradMA to accommodate FL with a large scale of workers. We then analyze convergence of GradMA theoretically under the smooth non-convex setting and show that its convergence rate achieves a linear speed up w.r.t the increasing number of sampled active workers. At last, our extensive experiments on various image classification tasks show that GradMA achieves significant performance gains in accuracy and communication efficiency compared to SOTA baselines.
♻ ☆ Safe Testing
We develop the theory of hypothesis testing based on the e-value, a notion of evidence that, unlike the p-value, allows for effortlessly combining results from several studies in the common scenario where the decision to perform a new study may depend on previous outcomes. Tests based on e-values are safe, i.e. they preserve Type-I error guarantees, under such optional continuation. We define growth-rate optimality (GRO) as an analogue of power in an optional continuation context, and we show how to construct GRO e-variables for general testing problems with composite null and alternative, emphasizing models with nuisance parameters. GRO e-values take the form of Bayes factors with special priors. We illustrate the theory using several classic examples including a one-sample safe t-test and the 2 x 2 contingency table. Sharing Fisherian, Neymanian and Jeffreys-Bayesian interpretations, e-values may provide a methodology acceptable to adherents of all three schools.
comment: Accepted as discussion paper to the Journal of the Royal Statistical Society series B
♻ ☆ Seq2Seq Surrogates of Epidemic Models to Facilitate Bayesian Inference
Epidemic models are powerful tools in understanding infectious disease. However, as they increase in size and complexity, they can quickly become computationally intractable. Recent progress in modelling methodology has shown that surrogate models can be used to emulate complex epidemic models with a high-dimensional parameter space. We show that deep sequence-to-sequence (seq2seq) models can serve as accurate surrogates for complex epidemic models with sequence based model parameters, effectively replicating seasonal and long-term transmission dynamics. Once trained, our surrogate can predict scenarios a several thousand times faster than the original model, making them ideal for policy exploration. We demonstrate that replacing a traditional epidemic model with a learned simulator facilitates robust Bayesian inference.
♻ ☆ Skew Class-balanced Re-weighting for Unbiased Scene Graph Generation
An unbiased scene graph generation (SGG) algorithm referred to as Skew Class-balanced Re-weighting (SCR) is proposed for considering the unbiased predicate prediction caused by the long-tailed distribution. The prior works focus mainly on alleviating the deteriorating performances of the minority predicate predictions, showing drastic dropping recall scores, i.e., losing the majority predicate performances. It has not yet correctly analyzed the trade-off between majority and minority predicate performances in the limited SGG datasets. In this paper, to alleviate the issue, the Skew Class-balanced Re-weighting (SCR) loss function is considered for the unbiased SGG models. Leveraged by the skewness of biased predicate predictions, the SCR estimates the target predicate weight coefficient and then re-weights more to the biased predicates for better trading-off between the majority predicates and the minority ones. Extensive experiments conducted on the standard Visual Genome dataset and Open Image V4 \& V6 show the performances and generality of the SCR with the traditional SGG models.
♻ ☆ Zero-One Laws of Graph Neural Networks
Graph neural networks (GNNs) are de facto standard deep learning architectures for machine learning on graphs. This has led to a large body of work analyzing the capabilities and limitations of these models, particularly pertaining to their representation and extrapolation capacity. We offer a novel theoretical perspective on the representation and extrapolation capacity of GNNs, by answering the question: how do GNNs behave as the number of graph nodes become very large? Under mild assumptions, we show that when we draw graphs of increasing size from the Erd\H{o}s-R\'enyi model, the probability that such graphs are mapped to a particular output by a class of GNN classifiers tends to either zero or to one. This class includes the popular graph convolutional network architecture. The result establishes 'zero-one laws' for these GNNs, and analogously to other convergence laws, entails theoretical limitations on their capacity. We empirically verify our results, observing that the theoretical asymptotic limits are evident already on relatively small graphs.
comment: 8 pages + references + 9 pages appendices, 2 figures
♻ ☆ A novel notion of barycenter for probability distributions based on optimal weak mass transport
We introduce weak barycenters of a family of probability distributions, based on the recently developed notion of optimal weak transport of mass by Gozlanet al. (2017) and Backhoff-Veraguas et al. (2020). We provide a theoretical analysis of this object and discuss its interpretation in the light of convex ordering between probability measures. In particular, we show that, rather than averaging the input distributions in a geometric way (as the Wasserstein barycenter based on classic optimal transport does) weak barycenters extract common geometric information shared by all the input distributions, encoded as a latent random variable that underlies all of them. We also provide an iterative algorithm to compute a weak barycenter for a finite family of input distributions, and a stochastic algorithm that computes them for arbitrary populations of laws. The latter approach is particularly well suited for the streaming setting, i.e., when distributions are observed sequentially. The notion of weak barycenter and our approaches to compute it are illustrated on synthetic examples, validated on 2D real-world data and compared to standard Wasserstein barycenters.
♻ ☆ SHINE: SHaring the INverse Estimate from the forward pass for bi-level optimization and implicit models ICLR 2022
In recent years, implicit deep learning has emerged as a method to increase the effective depth of deep neural networks. While their training is memory-efficient, they are still significantly slower to train than their explicit counterparts. In Deep Equilibrium Models (DEQs), the training is performed as a bi-level problem, and its computational complexity is partially driven by the iterative inversion of a huge Jacobian matrix. In this paper, we propose a novel strategy to tackle this computational bottleneck from which many bi-level problems suffer. The main idea is to use the quasi-Newton matrices from the forward pass to efficiently approximate the inverse Jacobian matrix in the direction needed for the gradient computation. We provide a theorem that motivates using our method with the original forward algorithms. In addition, by modifying these forward algorithms, we further provide theoretical guarantees that our method asymptotically estimates the true implicit gradient. We empirically study this approach and the recent Jacobian-Free method in different settings, ranging from hyperparameter optimization to large Multiscale DEQs (MDEQs) applied to CIFAR and ImageNet. Both methods reduce significantly the computational cost of the backward pass. While SHINE has a clear advantage on hyperparameter optimization problems, both methods attain similar computational performances for larger scale problems such as MDEQs at the cost of a limited performance drop compared to the original models.
comment: Accepted as a spotlight to ICLR 2022
♻ ☆ Exploiting Proximity-Aware Tasks for Embodied Social Navigation
Learning how to navigate among humans in an occluded and spatially constrained indoor environment, is a key ability required to embodied agent to be integrated into our society. In this paper, we propose an end-to-end architecture that exploits Proximity-Aware Tasks (referred as to Risk and Proximity Compass) to inject into a reinforcement learning navigation policy the ability to infer common-sense social behaviors. To this end, our tasks exploit the notion of immediate and future dangers of collision. Furthermore, we propose an evaluation protocol specifically designed for the Social Navigation Task in simulated environments. This is done to capture fine-grained features and characteristics of the policy by analyzing the minimal unit of human-robot spatial interaction, called Encounter. We validate our approach on Gibson4+ and Habitat-Matterport3D datasets.
♻ ☆ EiX-GNN : Concept-level eigencentrality explainer for graph neural networks
Nowadays, deep prediction models, especially graph neural networks, have a majorplace in critical applications. In such context, those models need to be highlyinterpretable or being explainable by humans, and at the societal scope, this understandingmay also be feasible for humans that do not have a strong prior knowledgein models and contexts that need to be explained. In the literature, explainingis a human knowledge transfer process regarding a phenomenon between an explainerand an explainee. We propose EiX-GNN (Eigencentrality eXplainer forGraph Neural Networks) a new powerful method for explaining graph neural networksthat encodes computationally this social explainer-to-explainee dependenceunderlying in the explanation process. To handle this dependency, we introducethe notion of explainee concept assimibility which allows explainer to adapt itsexplanation to explainee background or expectation. We lead a qualitative studyto illustrate our explainee concept assimibility notion on real-world data as wellas a qualitative study that compares, according to objective metrics established inthe literature, fairness and compactness of our method with respect to performingstate-of-the-art methods. It turns out that our method achieves strong results inboth aspects.
♻ ☆ Exploring the Relationship between Architecture and Adversarially Robust Generalization
Adversarial training has been demonstrated to be one of the most effective remedies for defending adversarial examples, yet it often suffers from the huge robustness generalization gap on unseen testing adversaries, deemed as the adversarially robust generalization problem. Despite the preliminary understandings devoted to adversarially robust generalization, little is known from the architectural perspective. To bridge the gap, this paper for the first time systematically investigated the relationship between adversarially robust generalization and architectural design. Inparticular, we comprehensively evaluated 20 most representative adversarially trained architectures on ImageNette and CIFAR-10 datasets towards multiple `p-norm adversarial attacks. Based on the extensive experiments, we found that, under aligned settings, Vision Transformers (e.g., PVT, CoAtNet) often yield better adversarially robust generalization while CNNs tend to overfit on specific attacks and fail to generalize on multiple adversaries. To better understand the nature behind it, we conduct theoretical analysis via the lens of Rademacher complexity. We revealed the fact that the higher weight sparsity contributes significantly towards the better adversarially robust generalization of Transformers, which can be often achieved by the specially-designed attention blocks. We hope our paper could help to better understand the mechanism for designing robust DNNs. Our model weights can be found at http://robust.art.
♻ ☆ No Reason for No Supervision: Improved Generalization in Supervised Models ICLR 2023
We consider the problem of training a deep neural network on a given classification task, e.g., ImageNet-1K (IN1K), so that it excels at both the training task as well as at other (future) transfer tasks. These two seemingly contradictory properties impose a trade-off between improving the model's generalization and maintaining its performance on the original task. Models trained with self-supervised learning tend to generalize better than their supervised counterparts for transfer learning; yet, they still lag behind supervised models on IN1K. In this paper, we propose a supervised learning setup that leverages the best of both worlds. We extensively analyze supervised training using multi-scale crops for data augmentation and an expendable projector head, and reveal that the design of the projector allows us to control the trade-off between performance on the training task and transferability. We further replace the last layer of class weights with class prototypes computed on the fly using a memory bank and derive two models: t-ReX that achieves a new state of the art for transfer learning and outperforms top methods such as DINO and PAWS on IN1K, and t-ReX* that matches the highly optimized RSB-A1 model on IN1K while performing better on transfer tasks. Code and pretrained models: https://europe.naverlabs.com/t-rex
comment: Accepted to ICLR 2023 (spotlight)
♻ ☆ The CMA Evolution Strategy: A Tutorial
This tutorial introduces the CMA Evolution Strategy (ES), where CMA stands for Covariance Matrix Adaptation. The CMA-ES is a stochastic, or randomized, method for real-parameter (continuous domain) optimization of non-linear, non-convex functions. We try to motivate and derive the algorithm from intuitive concepts and from requirements of non-linear, non-convex search in continuous domain.
comment: ArXiv e-prints, arXiv:1604.00772, 2016, pp.1-39
♻ ☆ Multivariate Probabilistic Forecasting of Intraday Electricity Prices using Normalizing Flows
Electricity is traded on various markets with different time horizons and regulations. Short-term intraday trading becomes increasingly important due to the higher penetration of renewables. In Germany, the intraday electricity price typically fluctuates around the day-ahead price of the European Power EXchange (EPEX) spot markets in a distinct hourly pattern. This work proposes a probabilistic modeling approach that models the intraday price difference to the day-ahead contracts. The model captures the emerging hourly pattern by considering the four 15 min intervals in each day-ahead price interval as a four-dimensional joint probability distribution. The resulting nontrivial, multivariate price difference distribution is learned using a normalizing flow, i.e., a deep generative model that combines conditional multivariate density estimation and probabilistic regression. Furthermore, this work discusses the influence of different external impact factors based on literature insights and impact analysis using explainable artificial intelligence (XAI). The normalizing flow is compared to an informed selection of historical data and probabilistic forecasts using a Gaussian copula and a Gaussian regression model. Among the different models, the normalizing flow identifies the trends with the highest accuracy and has the narrowest prediction intervals. Both the XAI analysis and the empirical experiments highlight that the immediate history of the price difference realization and the increments of the day-ahead price have the most substantial impact on the price difference.
comment: manuscript (20 pages, 11 figures, 5 tables), supporting information (8 pages, 5 figures, 4 tables)
♻ ☆ Advancing Spiking Neural Networks towards Deep Residual Learning
Despite the rapid progress of neuromorphic computing, inadequate capacity and insufficient representation power of spiking neural networks (SNNs) severely restrict their application scope in practice. Residual learning and shortcuts have been evidenced as an important approach for training deep neural networks, but rarely did previous work assess their applicability to the characteristics of spike-based communication and spatiotemporal dynamics. In this paper, we first identify that this negligence leads to impeded information flow and the accompanying degradation problem in previous residual SNNs. To address this issue, we propose a novel SNN-oriented residual architecture termed MS-ResNet, which establishes membrane-based shortcut pathways, and further prove that the gradient norm equality can be achieved in MS-ResNet by introducing block dynamical isometry theory, which ensures the network can be well-behaved in a depth-insensitive way. Thus we are able to significantly extend the depth of directly trained SNNs, e.g., up to 482 layers on CIFAR-10 and 104 layers on ImageNet, without observing any slight degradation problem. To validate the effectiveness of MS-ResNet, experiments on both frame-based and neuromorphic datasets are conducted. MS-ResNet104 achieves a superior result of 76.02% accuracy on ImageNet, which is the highest to our best knowledge in the domain of directly trained SNNs. Great energy efficiency is also observed, with an average of only one spike per neuron needed to classify an input sample. We believe our powerful and scalable models will provide a strong support for further exploration of SNNs.
♻ ☆ Privacy-Preserving and Lossless Distributed Estimation of High-Dimensional Generalized Additive Mixed Models
Various privacy-preserving frameworks that respect the individual's privacy in the analysis of data have been developed in recent years. However, available model classes such as simple statistics or generalized linear models lack the flexibility required for a good approximation of the underlying data-generating process in practice. In this paper, we propose an algorithm for a distributed, privacy-preserving, and lossless estimation of generalized additive mixed models (GAMM) using component-wise gradient boosting (CWB). Making use of CWB allows us to reframe the GAMM estimation as a distributed fitting of base learners using the $L_2$-loss. In order to account for the heterogeneity of different data location sites, we propose a distributed version of a row-wise tensor product that allows the computation of site-specific (smooth) effects. Our adaption of CWB preserves all the important properties of the original algorithm, such as an unbiased feature selection and the feasibility to fit models in high-dimensional feature spaces, and yields equivalent model estimates as CWB on pooled data. Next to a derivation of the equivalence of both algorithms, we also showcase the efficacy of our algorithm on a distributed heart disease data set and compare it with state-of-the-art methods.
♻ ☆ On the Sample Complexity of Two-Layer Networks: Lipschitz vs. Element-Wise Lipschitz Activation
We investigate the sample complexity of bounded two-layer neural networks using different activation functions. In particular, we consider the class $$ \mathcal{H} = \left\{\textbf{x}\mapsto \langle \textbf{v}, \sigma \circ W\textbf{b} + \textbf{b} \rangle : \textbf{b}\in\mathbb{R}^d, W \in \mathbb{R}^{\mathcal{T}\times d}, \textbf{v} \in \mathbb{R}^{\mathcal{T}}\right\} $$ where the spectral norm of $W$ and $\textbf{v}$ is bounded by $O(1)$, the Frobenius norm of $W$ is bounded from its initialization by $R > 0$, and $\sigma$ is a Lipschitz activation function. We prove that if $\sigma$ is element-wise, then the sample complexity of $\mathcal{H}$ has only logarithmic dependency in width and that this complexity is tight, up to logarithmic factors. We further show that the element-wise property of $\sigma$ is essential for a logarithmic dependency bound in width, in the sense that there exist non-element-wise activation functions whose sample complexity is linear in width, for widths that can be up to exponential in the input dimension. For the upper bound, we use the recent approach for norm-based bounds named Approximate Description Length (ADL) by arXiv:1910.05697. We further develop new techniques and tools for this approach that will hopefully inspire future works.
comment: 9 pages with additional 15 pages of supplementary
♻ ☆ Pistol: Pupil Invisible Supportive Tool to extract Pupil, Iris, Eye Opening, Eye Movements, Pupil and Iris Gaze Vector, and 2D as well as 3D Gaze
This paper describes a feature extraction and gaze estimation software, named \textit{Pistol} that can be used with Pupil Invisible projects and other eye trackers in the future. In offline mode, our software extracts multiple features from the eye including, the pupil and iris ellipse, eye aperture, pupil vector, iris vector, eye movement types from pupil and iris velocities, marker detection, marker distance, 2D gaze estimation for the pupil center, iris center, pupil vector, and iris vector using Levenberg Marquart fitting and neural networks. The gaze signal is computed in 2D for each eye and each feature separately and for both eyes in 3D also for each feature separately. We hope this software helps other researchers to extract state-of-the-art features for their research out of their recordings. Link: https://es-cloud.cs.uni-tuebingen.de/d/8e2ab8c3fdd444e1a135/?p=%2FPISTOL&mode=list
♻ ☆ Machine Learning Security in Industry: A Quantitative Survey
Despite the large body of academic work on machine learning security, little is known about the occurrence of attacks on machine learning systems in the wild. In this paper, we report on a quantitative study with 139 industrial practitioners. We analyze attack occurrence and concern and evaluate statistical hypotheses on factors influencing threat perception and exposure. Our results shed light on real-world attacks on deployed machine learning. On the organizational level, while we find no predictors for threat exposure in our sample, the amount of implement defenses depends on exposure to threats or expected likelihood to become a target. We also provide a detailed analysis of practitioners' replies on the relevance of individual machine learning attacks, unveiling complex concerns like unreliable decision making, business information leakage, and bias introduction into models. Finally, we find that on the individual level, prior knowledge about machine learning security influences threat perception. Our work paves the way for more research about adversarial machine learning in practice, but yields also insights for regulation and auditing.
comment: Accepted at TIFS, version with more detailed appendix containing more detailed statistical results. 17 pages, 6 tables and 4 figures
♻ ☆ You Only Need End-to-End Training for Long-Tailed Recognition
The generalization gap on the long-tailed data sets is largely owing to most categories only occupying a few training samples. Decoupled training achieves better performance by training backbone and classifier separately. What causes the poorer performance of end-to-end model training (e.g., logits margin-based methods)? In this work, we identify a key factor that affects the learning of the classifier: the channel-correlated features with low entropy before inputting into the classifier. From the perspective of information theory, we analyze why cross-entropy loss tends to produce highly correlated features on the imbalanced data. In addition, we theoretically analyze and prove its impacts on the gradients of classifier weights, the condition number of Hessian, and logits margin-based approach. Therefore, we firstly propose to use Channel Whitening to decorrelate ("scatter") the classifier's inputs for decoupling the weight update and reshaping the skewed decision boundary, which achieves satisfactory results combined with logits margin-based method. However, when the number of minor classes are large, batch imbalance and more participation in training cause over-fitting of the major classes. We also propose two novel modules, Block-based Relatively Balanced Batch Sampler (B3RS) and Batch Embedded Training (BET) to solve the above problems, which makes the end-to-end training achieve even better performance than decoupled training. Experimental results on the long-tailed classification benchmarks, CIFAR-LT and ImageNet-LT, demonstrate the effectiveness of our method.
comment: This is a draft
♻ ☆ Data-Efficient Training of CNNs and Transformers with Coresets: A Stability Perspective
Coreset selection is among the most effective ways to reduce the training time of CNNs, however, only limited is known on how the resultant models will behave under variations of the coreset size, and choice of datasets and models. Moreover, given the recent paradigm shift towards transformer-based models, it is still an open question how coreset selection would impact their performance. There are several similar intriguing questions that need to be answered for a wide acceptance of coreset selection methods, and this paper attempts to answer some of these. We present a systematic benchmarking setup and perform a rigorous comparison of different coreset selection methods on CNNs and transformers. Our investigation reveals that under certain circumstances, random selection of subsets is more robust and stable when compared with the SOTA selection methods. We demonstrate that the conventional concept of uniform subset sampling across the various classes of the data is not the appropriate choice. Rather samples should be adaptively chosen based on the complexity of the data distribution for each class. Transformers are generally pretrained on large datasets, and we show that for certain target datasets, it helps to keep their performance stable at even very small coreset sizes. We further show that when no pretraining is done or when the pretrained transformer models are used with non-natural images (e.g. medical data), CNNs tend to generalize better than transformers at even very small coreset sizes. Lastly, we demonstrate that in the absence of the right pretraining, CNNs are better at learning the semantic coherence between spatially distant objects within an image, and these tend to outperform transformers at almost all choices of the coreset size.
♻ ☆ Evolutionary Reinforcement Learning: A Survey
Reinforcement learning (RL) is a machine learning approach that trains agents to maximize cumulative rewards through interactions with environments. The integration of RL with deep learning has recently resulted in impressive achievements in a wide range of challenging tasks, including board games, arcade games, and robot control. Despite these successes, there remain several crucial challenges, including brittle convergence properties caused by sensitive hyperparameters, difficulties in temporal credit assignment with long time horizons and sparse rewards, a lack of diverse exploration, especially in continuous search space scenarios, difficulties in credit assignment in multi-agent reinforcement learning, and conflicting objectives for rewards. Evolutionary computation (EC), which maintains a population of learning agents, has demonstrated promising performance in addressing these limitations. This article presents a comprehensive survey of state-of-the-art methods for integrating EC into RL, referred to as evolutionary reinforcement learning (EvoRL). We categorize EvoRL methods according to key research fields in RL, including hyperparameter optimization, policy search, exploration, reward shaping, meta-RL, and multi-objective RL. We then discuss future research directions in terms of efficient methods, benchmarks, and scalable platforms. This survey serves as a resource for researchers and practitioners interested in the field of EvoRL, highlighting the important challenges and opportunities for future research. With the help of this survey, researchers and practitioners can develop more efficient methods and tailored benchmarks for EvoRL, further advancing this promising cross-disciplinary research field.
♻ ☆ DM-NeRF: 3D Scene Geometry Decomposition and Manipulation from 2D Images ICLR 2023
In this paper, we study the problem of 3D scene geometry decomposition and manipulation from 2D views. By leveraging the recent implicit neural representation techniques, particularly the appealing neural radiance fields, we introduce an object field component to learn unique codes for all individual objects in 3D space only from 2D supervision. The key to this component is a series of carefully designed loss functions to enable every 3D point, especially in non-occupied space, to be effectively optimized even without 3D labels. In addition, we introduce an inverse query algorithm to freely manipulate any specified 3D object shape in the learned scene representation. Notably, our manipulation algorithm can explicitly tackle key issues such as object collisions and visual occlusions. Our method, called DM-NeRF, is among the first to simultaneously reconstruct, decompose, manipulate and render complex 3D scenes in a single pipeline. Extensive experiments on three datasets clearly show that our method can accurately decompose all 3D objects from 2D views, allowing any interested object to be freely manipulated in 3D space such as translation, rotation, size adjustment, and deformation.
comment: ICLR 2023. Our data and code are available at: https://github.com/vLAR-group/DM-NeRF
♻ ☆ GFlowCausal: Generative Flow Networks for Causal Discovery
Causal discovery aims to uncover causal structure among a set of variables. Score-based approaches mainly focus on searching for the best Directed Acyclic Graph (DAG) based on a predefined score function. However, most of them are not applicable on a large scale due to the limited searchability. Inspired by the active learning in generative flow networks, we propose a novel approach to learning a DAG from observational data called GFlowCausal. It converts the graph search problem to a generation problem, in which direct edges are added gradually. GFlowCausal aims to learn the best policy to generate high-reward DAGs by sequential actions with probabilities proportional to predefined rewards. We propose a plug-and-play module based on transitive closure to ensure efficient sampling. Theoretical analysis shows that this module could guarantee acyclicity properties effectively and the consistency between final states and fully-connected graphs. We conduct extensive experiments on both synthetic and real datasets, and results show the proposed approach to be superior and also performs well in a large-scale setting.
♻ ☆ Joint Optimization of Energy Consumption and Completion Time in Federated Learning
Federated Learning (FL) is an intriguing distributed machine learning approach due to its privacy-preserving characteristics. To balance the trade-off between energy and execution latency, and thus accommodate different demands and application scenarios, we formulate an optimization problem to minimize a weighted sum of total energy consumption and completion time through two weight parameters. The optimization variables include bandwidth, transmission power and CPU frequency of each device in the FL system, where all devices are linked to a base station and train a global model collaboratively. Through decomposing the non-convex optimization problem into two subproblems, we devise a resource allocation algorithm to determine the bandwidth allocation, transmission power, and CPU frequency for each participating device. We further present the convergence analysis and computational complexity of the proposed algorithm. Numerical results show that our proposed algorithm not only has better performance at different weight parameters (i.e., different demands) but also outperforms the state of the art.
comment: This paper appears in the Proceedings of IEEE International Conference on Distributed Computing Systems (ICDCS) 2022. Please feel free to contact us for questions or remarks
♻ ☆ A numerical approximation method for the Fisher-Rao distance between multivariate normal distributions
We present a simple method to approximate Rao's distance between multivariate normal distributions based on discretizing curves joining normal distributions and approximating Rao distances between successive nearby normal distributions on the curves by the square root of Jeffreys divergence. We consider experimentally the linear interpolation curves in the ordinary, natural and expectation parameterizations of the normal distributions, and compare these curves with a curve derived from the Calvo and Oller's isometric embedding of the Fisher-Rao $d$-variate normal manifold into the cone of $(d+1)\times (d+1)$ symmetric positive-definite matrices [Journal of multivariate analysis 35.2 (1990): 223-242]. We report on our experiments and assess the quality of our approximation technique by comparing the numerical approximations with lower and upper bounds. Finally, we present some information-geometric properties of the Calvo and Oller's isometric embedding.
comment: 30 pages, 16 figures, 3 tables
♻ ☆ ABAW: Valence-Arousal Estimation, Expression Recognition, Action Unit Detection & Emotional Reaction Intensity Estimation Challenges
The fifth Affective Behavior Analysis in-the-wild (ABAW) Competition is part of the respective ABAW Workshop which will be held in conjunction with IEEE Computer Vision and Pattern Recognition Conference (CVPR), 2023. The 5th ABAW Competition is a continuation of the Competitions held at ECCV 2022, IEEE CVPR 2022, ICCV 2021, IEEE FG 2020 and CVPR 2017 Conferences, and is dedicated at automatically analyzing affect. For this year's Competition, we feature two corpora: i) an extended version of the Aff-Wild2 database and ii) the Hume-Reaction dataset. The former database is an audiovisual one of around 600 videos of around 3M frames and is annotated with respect to:a) two continuous affect dimensions -valence (how positive/negative a person is) and arousal (how active/passive a person is)-; b) basic expressions (e.g. happiness, sadness, neutral state); and c) atomic facial muscle actions (i.e., action units). The latter dataset is an audiovisual one in which reactions of individuals to emotional stimuli have been annotated with respect to seven emotional expression intensities. Thus the 5th ABAW Competition encompasses four Challenges: i) uni-task Valence-Arousal Estimation, ii) uni-task Expression Classification, iii) uni-task Action Unit Detection, and iv) Emotional Reaction Intensity Estimation. In this paper, we present these Challenges, along with their corpora, we outline the evaluation metrics, we present the baseline systems and illustrate their obtained performance.
comment: arXiv admin note: text overlap with arXiv:2202.10659
♻ ☆ Combining Contention-Based Spectrum Access and Adaptive Modulation using Deep Reinforcement Learning
The use of unlicensed spectrum for cellular systems to mitigate spectrum scarcity has led to the development of intelligent adaptive approaches to spectrum access that improve upon traditional carrier sensing and listen-before-talk methods. We study decentralized contention-based medium access for base stations (BSs) of a single Radio Access Technology (RAT) operating on unlicensed shared spectrum. We devise a distributed deep reinforcement learning-based algorithm for both contention and adaptive modulation, modelled on a two state Markov decision process, that attempts to maximize a network-wide downlink throughput objective. Empirically, we find the (proportional fairness) reward accumulated by a policy gradient approach to be significantly higher than even a genie-aided adaptive energy detection threshold. Our approaches are further validated by improved sum and peak throughput. The scalability of our approach to large networks is demonstrated via an improved cumulative reward earned on both indoor and outdoor layouts with a large number of BSs.
comment: 6 pages, 3 figures. Published in Asilomar 2022
♻ ☆ Multi-Task Recommendations with Reinforcement Learning
In recent years, Multi-task Learning (MTL) has yielded immense success in Recommender System (RS) applications. However, current MTL-based recommendation models tend to disregard the session-wise patterns of user-item interactions because they are predominantly constructed based on item-wise datasets. Moreover, balancing multiple objectives has always been a challenge in this field, which is typically avoided via linear estimations in existing works. To address these issues, in this paper, we propose a Reinforcement Learning (RL) enhanced MTL framework, namely RMTL, to combine the losses of different recommendation tasks using dynamic weights. To be specific, the RMTL structure can address the two aforementioned issues by (i) constructing an MTL environment from session-wise interactions and (ii) training multi-task actor-critic network structure, which is compatible with most existing MTL-based recommendation models, and (iii) optimizing and fine-tuning the MTL loss function using the weights generated by critic networks. Experiments on two real-world public datasets demonstrate the effectiveness of RMTL with a higher AUC against state-of-the-art MTL-based recommendation models. Additionally, we evaluate and validate RMTL's compatibility and transferability across various MTL models.
comment: TheWebConf2023
♻ ☆ RawNet: Fast End-to-End Neural Vocoder
Neural network-based vocoders have recently demonstrated the powerful ability to synthesize high-quality speech. These models usually generate samples by conditioning on spectral features, such as Mel-spectrogram and fundamental frequency, which is crucial to speech synthesis. However, the feature extraction procession tends to depend heavily on human knowledge resulting in a less expressive description of the origin audio. In this work, we proposed RawNet, a complete end-to-end neural vocoder following the auto-encoder structure for speaker-dependent and -independent speech synthesis. It automatically learns to extract features and recover audio using neural networks, which include a coder network to capture a higher representation of the input audio and an autoregressive voder network to restore the audio in a sample-by-sample manner. The coder and voder are jointly trained directly on the raw waveform without any human-designed features. The experimental results show that RawNet achieves a better speech quality using a simplified model architecture and obtains a faster speech generation speed at the inference stage.
♻ ☆ Diagnosing Model Performance Under Distribution Shift
Prediction models can perform poorly when deployed to target distributions different from the training distribution. To understand these operational failure modes, we develop a method, called DIstribution Shift DEcomposition (DISDE), to attribute a drop in performance to different types of distribution shifts. Our approach decomposes the performance drop into terms for 1) an increase in harder but frequently seen examples from training, 2) changes in the relationship between features and outcomes, and 3) poor performance on examples infrequent or unseen during training. These terms are defined by fixing a distribution on $X$ while varying the conditional distribution of $Y \mid X$ between training and target, or by fixing the conditional distribution of $Y \mid X$ while varying the distribution on $X$. In order to do this, we define a hypothetical distribution on $X$ consisting of values common in both training and target, over which it is easy to compare $Y \mid X$ and thus predictive performance. We estimate performance on this hypothetical distribution via reweighting methods. Empirically, we show how our method can 1) inform potential modeling improvements across distribution shifts for employment prediction on tabular census data, and 2) help to explain why certain domain adaptation methods fail to improve model performance for satellite image classification.
♻ ☆ Reproducible and Portable Big Data Analytics in the Cloud
Cloud computing has become a major approach to help reproduce computational experiments. Yet there are still two main difficulties in reproducing batch based big data analytics (including descriptive and predictive analytics) in the cloud. The first is how to automate end-to-end scalable execution of analytics including distributed environment provisioning, analytics pipeline description, parallel execution, and resource termination. The second is that an application developed for one cloud is difficult to be reproduced in another cloud, a.k.a. vendor lock-in problem. To tackle these problems, we leverage serverless computing and containerization techniques for automated scalable execution and reproducibility, and utilize the adapter design pattern to enable application portability and reproducibility across different clouds. We propose and develop an open-source toolkit that supports 1) fully automated end-to-end execution and reproduction via a single command, 2) automated data and configuration storage for each execution, 3) flexible client modes based on user preferences, 4) execution history query, and 5) simple reproduction of existing executions in the same environment or a different environment. We did extensive experiments on both AWS and Azure using four big data analytics applications that run on virtual CPU/GPU clusters. The experiments show our toolkit can achieve good execution performance, scalability, and efficient reproducibility for cloud-based big data analytics.
comment: accepted by journal IEEE Transactions on Cloud Computing
♻ ☆ CoSyn: Detecting Implicit Hate Speech in Online Conversations Using a Context Synergized Hyperbolic Network IJCAI 2023
The tremendous growth of social media users interacting in online conversations has also led to significant growth in hate speech. Most of the prior works focus on detecting explicit hate speech, which is overt and leverages hateful phrases, with very little work focusing on detecting hate speech that is implicit or denotes hatred through indirect or coded language. In this paper, we present CoSyn, a user- and conversational-context synergized network for detecting implicit hate speech in online conversation trees. CoSyn first models the user's personal historical and social context using a novel hyperbolic Fourier attention mechanism and hyperbolic graph convolution network. Next, we jointly model the user's personal context and the conversational context using a novel context interaction mechanism in the hyperbolic space that clearly captures the interplay between the two and makes independent assessments on the amounts of information to be retrieved from both contexts. CoSyn performs all operations in the hyperbolic space to account for the scale-free dynamics of social media. We demonstrate the effectiveness of CoSyn both qualitatively and quantitatively on an open-source hate speech dataset with Twitter conversations and show that CoSyn outperforms all our baselines in detecting implicit hate speech with absolute improvements in the range of 8.15% - 19.50%.
comment: Under review at IJCAI 2023
♻ ☆ ReAct: Synergizing Reasoning and Acting in Language Models ICLR
While large language models (LLMs) have demonstrated impressive capabilities across tasks in language understanding and interactive decision making, their abilities for reasoning (e.g. chain-of-thought prompting) and acting (e.g. action plan generation) have primarily been studied as separate topics. In this paper, we explore the use of LLMs to generate both reasoning traces and task-specific actions in an interleaved manner, allowing for greater synergy between the two: reasoning traces help the model induce, track, and update action plans as well as handle exceptions, while actions allow it to interface with external sources, such as knowledge bases or environments, to gather additional information. We apply our approach, named ReAct, to a diverse set of language and decision making tasks and demonstrate its effectiveness over state-of-the-art baselines, as well as improved human interpretability and trustworthiness over methods without reasoning or acting components. Concretely, on question answering (HotpotQA) and fact verification (Fever), ReAct overcomes issues of hallucination and error propagation prevalent in chain-of-thought reasoning by interacting with a simple Wikipedia API, and generates human-like task-solving trajectories that are more interpretable than baselines without reasoning traces. On two interactive decision making benchmarks (ALFWorld and WebShop), ReAct outperforms imitation and reinforcement learning methods by an absolute success rate of 34% and 10% respectively, while being prompted with only one or two in-context examples. Project site with code: https://react-lm.github.io
comment: v3 is the ICLR camera ready version with some typos fixed. Project site with code: https://react-lm.github.io
Multimedia
☆ QVRF: A Quantization-error-aware Variable Rate Framework for Learned Image Compression
Learned image compression has exhibited promising compression performance, but variable bitrates over a wide range remain a challenge. State-of-the-art variable rate methods compromise the loss of model performance and require numerous additional parameters. In this paper, we present a Quantization-error-aware Variable Rate Framework (QVRF) that utilizes a univariate quantization regulator a to achieve wide-range variable rates within a single model. Specifically, QVRF defines a quantization regulator vector coupled with predefined Lagrange multipliers to control quantization error of all latent representation for discrete variable rates. Additionally, the reparameterization method makes QVRF compatible with a round quantizer. Exhaustive experiments demonstrate that existing fixed-rate VAE-based methods equipped with QVRF can achieve wide-range continuous variable rates within a single model without significant performance degradation. Furthermore, QVRF outperforms contemporary variable-rate methods in rate-distortion performance with minimal additional parameters.
comment: 7 pages, 6 figures
☆ MuLTI: Efficient Video-and-Language Understanding with MultiWay-Sampler and Multiple Choice Modeling
Video-and-language understanding has a variety of applications in the industry, such as video question answering, text-video retrieval and multi-label classification. Existing video-and-language understanding methods generally adopt heavy multi-modal encoders and feature fusion modules, which consume large amounts of GPU memory. Especially, they have difficulty dealing with dense video frames or long text that are prevalent in industrial applications. In this paper, we propose MuLTI, a highly accurate and memory-efficient video-and-language understanding model that achieves efficient and effective feature fusion through feature sampling and attention modules. Therefore, MuLTI can handle longer sequences with limited GPU memory. Then, we introduce an attention-based adapter to the encoders, which finetunes the shallow features to improve the model's performance with low GPU memory consumption. Finally, to further improve the model's performance, we introduce a new pretraining task named Multiple Choice Modeling to bridge the task gap between pretraining and downstream tasks and enhance the model's ability to align the video and the text. Benefiting from the efficient feature fusion module, the attention-based adapter and the new pretraining task, MuLTI achieves state-of-the-art performance on multiple datasets. Implementation and pretrained models will be released.
♻ ☆ Neural Video Compression with Diverse Contexts CVPR 2023
For any video codecs, the coding efficiency highly relies on whether the current signal to be encoded can find the relevant contexts from the previous reconstructed signals. Traditional codec has verified more contexts bring substantial coding gain, but in a time-consuming manner. However, for the emerging neural video codec (NVC), its contexts are still limited, leading to low compression ratio. To boost NVC, this paper proposes increasing the context diversity in both temporal and spatial dimensions. First, we guide the model to learn hierarchical quality patterns across frames, which enriches long-term and yet high-quality temporal contexts. Furthermore, to tap the potential of optical flow-based coding framework, we introduce a group-based offset diversity where the cross-group interaction is proposed for better context mining. In addition, this paper also adopts a quadtree-based partition to increase spatial context diversity when encoding the latent representation in parallel. Experiments show that our codec obtains 23.5% bitrate saving over previous SOTA NVC. Better yet, our codec has surpassed the under-developing next generation traditional codec/ECM in both RGB and YUV420 colorspaces, in terms of PSNR. The codes are at https://github.com/microsoft/DCVC.
comment: Accepted by CVPR 2023. Codes are at https://github.com/microsoft/DCVC
♻ ☆ Contrastive Audio-Visual Masked Autoencoder ICLR 2023
In this paper, we first extend the recent Masked Auto-Encoder (MAE) model from a single modality to audio-visual multi-modalities. Subsequently, we propose the Contrastive Audio-Visual Masked Auto-Encoder (CAV-MAE) by combining contrastive learning and masked data modeling, two major self-supervised learning frameworks, to learn a joint and coordinated audio-visual representation. Our experiments show that the contrastive audio-visual correspondence learning objective not only enables the model to perform audio-visual retrieval tasks, but also helps the model learn a better joint representation. As a result, our fully self-supervised pretrained CAV-MAE achieves a new SOTA accuracy of 65.9% on VGGSound, and is comparable with the previous best supervised pretrained model on AudioSet in the audio-visual event classification task. Code and pretrained models are at https://github.com/yuangongnd/cav-mae.
comment: Accepted at ICLR 2023 as a notable top 25% paper. Code and pretrained models are at https://github.com/yuangongnd/cav-mae
Computation and Language
☆ Planning with Large Language Models for Code Generation ICLR 2023
Existing large language model-based code generation pipelines typically use beam search or sampling algorithms during the decoding process. Although the programs they generate achieve high token-matching-based scores, they often fail to compile or generate incorrect outputs. The main reason is that conventional Transformer decoding algorithms may not be the best choice for code generation. In this work, we propose a novel Transformer decoding algorithm, Planning-Guided Transformer Decoding (PG-TD), that uses a planning algorithm to do lookahead search and guide the Transformer to generate better programs. Specifically, instead of simply optimizing the likelihood of the generated sequences, the Transformer makes use of a planner to generate candidate programs and test them on public test cases. The Transformer can therefore make more informed decisions and generate tokens that will eventually lead to higher-quality programs. We also design a mechanism that shares information between the Transformer and the planner to make our algorithm computationally efficient. We empirically evaluate our framework with several large language models as backbones on public coding challenge benchmarks, showing that 1) it can generate programs that consistently achieve higher performance compared with competing baseline methods; 2) it enables controllable code generation, such as concise codes and highly-commented codes by optimizing modified objective.
comment: ICLR 2023. Project page:https://codeaimcts.github.io
☆ Personalisation within bounds: A risk taxonomy and policy framework for the alignment of large language models with personalised feedback
Large language models (LLMs) are used to generate content for a wide range of tasks, and are set to reach a growing audience in coming years due to integration in product interfaces like ChatGPT or search engines like Bing. This intensifies the need to ensure that models are aligned with human preferences and do not produce unsafe, inaccurate or toxic outputs. While alignment techniques like reinforcement learning with human feedback (RLHF) and red-teaming can mitigate some safety concerns and improve model capabilities, it is unlikely that an aggregate fine-tuning process can adequately represent the full range of users' preferences and values. Different people may legitimately disagree on their preferences for language and conversational norms, as well as on values or ideologies which guide their communication. Personalising LLMs through micro-level preference learning processes may result in models that are better aligned with each user. However, there are several normative challenges in defining the bounds of a societally-acceptable and safe degree of personalisation. In this paper, we ask how, and in what ways, LLMs should be personalised. First, we review literature on current paradigms for aligning LLMs with human feedback, and identify issues including (i) a lack of clarity regarding what alignment means; (ii) a tendency of technology providers to prescribe definitions of inherently subjective preferences and values; and (iii) a 'tyranny of the crowdworker', exacerbated by a lack of documentation in who we are really aligning to. Second, we present a taxonomy of benefits and risks associated with personalised LLMs, for individuals and society at large. Finally, we propose a three-tiered policy framework that allows users to experience the benefits of personalised alignment, while restraining unsafe and undesirable LLM-behaviours within (supra-)national and organisational bounds.
comment: 19 pages, 1 table
☆ Seeing ChatGPT Through Students' Eyes: An Analysis of TikTok Data
Advanced large language models like ChatGPT have gained considerable attention recently, including among students. However, while the debate on ChatGPT in academia is making waves, more understanding is needed among lecturers and teachers on how students use and perceive ChatGPT. To address this gap, we analyzed the content on ChatGPT available on TikTok in February 2023. TikTok is a rapidly growing social media platform popular among individuals under 30. Specifically, we analyzed the content of the 100 most popular videos in English tagged with #chatgpt, which collectively garnered over 250 million views. Most of the videos we studied promoted the use of ChatGPT for tasks like writing essays or code. In addition, many videos discussed AI detectors, with a focus on how other tools can help to transform ChatGPT output to fool these detectors. This also mirrors the discussion among educators on how to treat ChatGPT as lecturers and teachers in teaching and grading. What is, however, missing from the analyzed clips on TikTok are videos that discuss ChatGPT producing content that is nonsensical or unfaithful to the training data.
☆ Replacement as a Self-supervision for Fine-grained Vision-language Pre-training
Fine-grained supervision based on object annotations has been widely used for vision and language pre-training (VLP). However, in real-world application scenarios, aligned multi-modal data is usually in the image-caption format, which only provides coarse-grained supervision. It is cost-expensive to collect object annotations and build object annotation pre-extractor for different scenarios. In this paper, we propose a fine-grained self-supervision signal without object annotations from a replacement perspective. First, we propose a homonym sentence rewriting (HSR) algorithm to provide token-level supervision. The algorithm replaces a verb/noun/adjective/quantifier word of the caption with its homonyms from WordNet. Correspondingly, we propose a replacement vision-language modeling (RVLM) framework to exploit the token-level supervision. Two replaced modeling tasks, i.e., replaced language contrastive (RLC) and replaced language modeling (RLM), are proposed to learn the fine-grained alignment. Extensive experiments on several downstream tasks demonstrate the superior performance of the proposed method.
comment: Work in progress
☆ MixSpeech: Cross-Modality Self-Learning with Audio-Visual Stream Mixup for Visual Speech Translation and Recognition
Multi-media communications facilitate global interaction among people. However, despite researchers exploring cross-lingual translation techniques such as machine translation and audio speech translation to overcome language barriers, there is still a shortage of cross-lingual studies on visual speech. This lack of research is mainly due to the absence of datasets containing visual speech and translated text pairs. In this paper, we present \textbf{AVMuST-TED}, the first dataset for \textbf{A}udio-\textbf{V}isual \textbf{Mu}ltilingual \textbf{S}peech \textbf{T}ranslation, derived from \textbf{TED} talks. Nonetheless, visual speech is not as distinguishable as audio speech, making it difficult to develop a mapping from source speech phonemes to the target language text. To address this issue, we propose MixSpeech, a cross-modality self-learning framework that utilizes audio speech to regularize the training of visual speech tasks. To further minimize the cross-modality gap and its impact on knowledge transfer, we suggest adopting mixed speech, which is created by interpolating audio and visual streams, along with a curriculum learning strategy to adjust the mixing ratio as needed. MixSpeech enhances speech translation in noisy environments, improving BLEU scores for four languages on AVMuST-TED by +1.4 to +4.2. Moreover, it achieves state-of-the-art performance in lip reading on CMLR (11.1\%), LRS2 (25.5\%), and LRS3 (28.0\%).
comment: https://github.com/Exgc/AVMuST-TED
☆ Dynamic Stashing Quantization for Efficient Transformer Training
Large Language Models (LLMs) have demonstrated impressive performance on a range of Natural Language Processing (NLP) tasks. Unfortunately, the immense amount of computations and memory accesses required for LLM training makes them prohibitively expensive in terms of hardware cost, and thus challenging to deploy in use cases such as on-device learning. In this paper, motivated by the observation that LLM training is memory-bound, we propose a novel dynamic quantization strategy, termed Dynamic Stashing Quantization (DSQ), that puts a special focus on reducing the memory operations, but also enjoys the other benefits of low precision training, such as the reduced arithmetic cost. We conduct a thorough study on two translation tasks (trained-from-scratch) and three classification tasks (fine-tuning). DSQ reduces the amount of arithmetic operations by $20.95\times$ and the number of DRAM operations by $2.55\times$ on IWSLT17 compared to the standard 16-bit fixed-point, which is widely used in on-device learning.
☆ SEAM: An Integrated Activation-Coupled Model of Sentence Processing and Eye Movements in Reading
Models of eye-movement control during reading, developed largely within psychology, usually focus on visual, attentional, and motor processes but neglect post-lexical language processing; by contrast, models of sentence comprehension processes, developed largely within psycholinguistics, generally focus only on post-lexical language processes. We present a model that combines these two research threads, by integrating eye-movement control and sentence processing. Developing such an integrated model is extremely challenging and computationally demanding, but such an integration is an important step toward complete mathematical models of natural language comprehension in reading. We combine the SWIFT model of eye-movement control (Engbert et al., Psychological Review, 112, 2005, pp. 777-813) with key components of the Lewis and Vasishth sentence processing model (Lewis and Vasishth, Cognitive Science, 29, 2005, pp. 375-419). This integration becomes possible, for the first time, due in part to recent advances in successful parameter identification in dynamical models, which allows us to investigate profile log-likelihoods for individual model parameters. We present a fully implemented proof-of-concept model demonstrating how such an integrated model can be achieved; our approach includes Bayesian model inference with Markov Chain Monte Carlo (MCMC) sampling as a key computational tool. The integrated model, SEAM, can successfully reproduce eye movement patterns that arise due to similarity-based interference in reading. To our knowledge, this is the first-ever integration of a complete process model of eye-movement control with linguistic dependency completion processes in sentence comprehension. In future work, this proof of concept model will need to be evaluated using a comprehensive set of benchmark data.
☆ Geometry of Language
In this article, we present a fresh perspective on language, combining ideas from various sources, but mixed in a new synthesis. As in the minimalist program, the question is whether we can formulate an elegant formalism, a universal grammar or a mechanism which explains significant aspects of the human faculty of language, which in turn can be considered a natural disposition for the evolution and deployment of the diverse human languages. We describe such a mechanism, which differs from existing logical and grammatical approaches by its geometric nature. Our main contribution is to explore the assumption that sentence recognition takes place by forming chains of tokens representing words, followed by matching these chains with pre-existing chains representing grammatical word orders. The aligned chains of tokens give rise to two- and three-dimensional complexes. The resulting model gives an alternative presentation for subtle rules, traditionally formalized using categorial grammar.
comment: 17 pages, 24 figures
☆ $π$-augmented pregroups and applications to linguistics
We enrich pregroups with a mapping which allows us to locally apply precyclic permutations to designated substrings. We prove a normalisation theorem for such algebraic structures and briefly formalise some known applications of pregroups to the analysis of clitic pronouns in certain natural languages.
comment: 20 pages
☆ Can a Frozen Pretrained Language Model be used for Zero-shot Neural Retrieval on Entity-centric Questions? AAAI 2023
Neural document retrievers, including dense passage retrieval (DPR), have outperformed classical lexical-matching retrievers, such as BM25, when fine-tuned and tested on specific question-answering datasets. However, it has been shown that the existing dense retrievers do not generalize well not only out of domain but even in domain such as Wikipedia, especially when a named entity in a question is a dominant clue for retrieval. In this paper, we propose an approach toward in-domain generalization using the embeddings generated by the frozen language model trained with the entities in the domain. By not fine-tuning, we explore the possibility that the rich knowledge contained in a pretrained language model can be used for retrieval tasks. The proposed method outperforms conventional DPRs on entity-centric questions in Wikipedia domain and achieves almost comparable performance to BM25 and state-of-the-art SPAR model. We also show that the contextualized keys lead to strong improvements compared to BM25 when the entity names consist of common words. Our results demonstrate the feasibility of the zero-shot retrieval method for entity-centric questions of Wikipedia domain, where DPR has struggled to perform.
comment: Accepted to Workshop on Knowledge Augmented Methods for Natural Language Processing, in conjunction with AAAI 2023
☆ ESCL: Equivariant Self-Contrastive Learning for Sentence Representations ICASSP 2023
Previous contrastive learning methods for sentence representations often focus on insensitive transformations to produce positive pairs, but neglect the role of sensitive transformations that are harmful to semantic representations. Therefore, we propose an Equivariant Self-Contrastive Learning (ESCL) method to make full use of sensitive transformations, which encourages the learned representations to be sensitive to certain types of transformations with an additional equivariant learning task. Meanwhile, in order to improve practicability and generality, ESCL simplifies the implementations of traditional equivariant contrastive methods to share model parameters from the perspective of multi-task learning. We evaluate our ESCL on semantic textual similarity tasks. The proposed method achieves better results while using fewer learning parameters compared to previous methods.
comment: accepted by ICASSP 2023
☆ Improving Video Retrieval by Adaptive Margin SIGIR 2021
Video retrieval is becoming increasingly important owing to the rapid emergence of videos on the Internet. The dominant paradigm for video retrieval learns video-text representations by pushing the distance between the similarity of positive pairs and that of negative pairs apart from a fixed margin. However, negative pairs used for training are sampled randomly, which indicates that the semantics between negative pairs may be related or even equivalent, while most methods still enforce dissimilar representations to decrease their similarity. This phenomenon leads to inaccurate supervision and poor performance in learning video-text representations. While most video retrieval methods overlook that phenomenon, we propose an adaptive margin changed with the distance between positive and negative pairs to solve the aforementioned issue. First, we design the calculation framework of the adaptive margin, including the method of distance measurement and the function between the distance and the margin. Then, we explore a novel implementation called "Cross-Modal Generalized Self-Distillation" (CMGSD), which can be built on the top of most video retrieval models with few modifications. Notably, CMGSD adds few computational overheads at train time and adds no computational overhead at test time. Experimental results on three widely used datasets demonstrate that the proposed method can yield significantly better performance than the corresponding backbone model, and it outperforms state-of-the-art methods by a large margin.
comment: Accepted by SIGIR 2021
☆ Dynamic Multi-View Fusion Mechanism For Chinese Relation Extraction PAKDD 2023
Recently, many studies incorporate external knowledge into character-level feature based models to improve the performance of Chinese relation extraction. However, these methods tend to ignore the internal information of the Chinese character and cannot filter out the noisy information of external knowledge. To address these issues, we propose a mixture-of-view-experts framework (MoVE) to dynamically learn multi-view features for Chinese relation extraction. With both the internal and external knowledge of Chinese characters, our framework can better capture the semantic information of Chinese characters. To demonstrate the effectiveness of the proposed framework, we conduct extensive experiments on three real-world datasets in distinct domains. Experimental results show consistent and significant superiority and robustness of our proposed framework. Our code and dataset will be released at: https://gitee.com/tmg-nudt/multi-view-of-expert-for-chineserelation-extraction
comment: This paper has been accepted by PAKDD 2023
☆ Revisiting the relevance of traditional genres: a network analysis of fiction readers' preferences
We investigate how well traditional fiction genres like Fantasy, Thriller, and Literature represent readers' preferences. Using user data from Goodreads we construct a book network where two books are strongly linked if the same people tend to read or enjoy them both. We then partition this network into communities of similar books and assign each a list of subjects from The Open Library to serve as a proxy for traditional genres. Our analysis reveals that the network communities correspond to existing combinations of traditional genres, but that the exact communities differ depending on whether we consider books that people read or books that people enjoy. In addition, we apply principal component analysis to the data and find that the variance in the book communities is best explained by two factors: the maturity/childishness and realism/fantastical nature of the books. We propose using this maturity-realism plane as a coarse classification tool for stories.
comment: Supplementary materials at https://github.com/taomsakal/book-networks
☆ Learning the Legibility of Visual Text Perturbations EACL 2023
Many adversarial attacks in NLP perturb inputs to produce visually similar strings ('ergo' $\rightarrow$ '$\epsilon$rgo') which are legible to humans but degrade model performance. Although preserving legibility is a necessary condition for text perturbation, little work has been done to systematically characterize it; instead, legibility is typically loosely enforced via intuitions around the nature and extent of perturbations. Particularly, it is unclear to what extent can inputs be perturbed while preserving legibility, or how to quantify the legibility of a perturbed string. In this work, we address this gap by learning models that predict the legibility of a perturbed string, and rank candidate perturbations based on their legibility. To do so, we collect and release \dataset, a human-annotated dataset comprising the legibility of visually perturbed text. Using this dataset, we build both text- and vision-based models which achieve up to $0.91$ F1 score in predicting whether an input is legible, and an accuracy of $0.86$ in predicting which of two given perturbations is more legible. Additionally, we discover that legible perturbations from the \dataset dataset are more effective at lowering the performance of NLP models than best-known attack strategies, suggesting that current models may be vulnerable to a broad range of perturbations beyond what is captured by existing visual attacks. Data, code, and models are available at https://github.com/dvsth/learning-legibility-2023.
comment: 9 pages, 11 figures. Long paper at EACL 2023
☆ ICL-D3IE: In-Context Learning with Diverse Demonstrations Updating for Document Information Extraction
Large language models (LLMs), such as GPT-3 and ChatGPT, have demonstrated remarkable results in various natural language processing (NLP) tasks with in-context learning, which involves inference based on a few demonstration examples. Despite their successes in NLP tasks, no investigation has been conducted to assess the ability of LLMs to perform document information extraction (DIE) using in-context learning. Applying LLMs to DIE poses two challenges: the modality and task gap. To this end, we propose a simple but effective in-context learning framework called ICL-D3IE, which enables LLMs to perform DIE with different types of demonstration examples. Specifically, we extract the most difficult and distinct segments from hard training documents as hard demonstrations for benefiting all test instances. We design demonstrations describing relationships that enable LLMs to understand positional relationships. We introduce formatting demonstrations for easy answer extraction. Additionally, the framework improves diverse demonstrations by updating them iteratively. Our experiments on three widely used benchmark datasets demonstrate that the ICL-D3IE framework enables GPT-3/ChatGPT to achieve superior performance when compared to previous pre-trained methods fine-tuned with full training in both the in-distribution (ID) setting and in the out-of-distribution (OOD) setting.
☆ Unsupervised Language agnostic WER Standardization
Word error rate (WER) is a standard metric for the evaluation of Automated Speech Recognition (ASR) systems. However, WER fails to provide a fair evaluation of human perceived quality in presence of spelling variations, abbreviations, or compound words arising out of agglutination. Multiple spelling variations might be acceptable based on locale/geography, alternative abbreviations, borrowed words, and transliteration of code-mixed words from a foreign language to the target language script. Similarly, in case of agglutination, often times the agglutinated, as well as the split forms, are acceptable. Previous work handled this problem by using manually identified normalization pairs and applying them to both the transcription and the hypothesis before computing WER. In this paper, we propose an automatic WER normalization system consisting of two modules: spelling normalization and segmentation normalization. The proposed system is unsupervised and language agnostic, and therefore scalable. Experiments with ASR on 35K utterances across four languages yielded an average WER reduction of 13.28%. Human judgements of these automatically identified normalization pairs show that our WER-normalized evaluation is highly consistent with the perceived quality of ASR output.
☆ Multi-Stage Coarse-to-Fine Contrastive Learning for Conversation Intent Induction
Intent recognition is critical for task-oriented dialogue systems. However, for emerging domains and new services, it is difficult to accurately identify the key intent of a conversation due to time-consuming data annotation and comparatively poor model transferability. Therefore, the automatic induction of dialogue intention is very important for intelligent dialogue systems. This paper presents our solution to Track 2 of Intent Induction from Conversations for Task-Oriented Dialogue at the Eleventh Dialogue System Technology Challenge (DSTC11). The essence of intention clustering lies in distinguishing the representation of different dialogue utterances. The key to automatic intention induction is that, for any given set of new data, the sentence representation obtained by the model can be well distinguished from different labels. Therefore, we propose a multi-stage coarse-to-fine contrastive learning model training scheme including unsupervised contrastive learning pre-training, supervised contrastive learning pre-training, and fine-tuning with joint contrastive learning and clustering to obtain a better dialogue utterance representation model for the clustering task. In the released DSTC11 Track 2 evaluation results, our proposed system ranked first on both of the two subtasks of this Track.
comment: Ranked 1st on Track 2 at DSTC 11, Accepted by DSTC 11 Workshop
☆ Let's Get Personal: Personal Questions Improve SocialBot Performance in the Alexa Prize
There has been an increased focus on creating conversational open-domain dialogue systems in the spoken dialogue community. Unlike traditional dialogue systems, these conversational systems cannot assume any specific information need or domain restrictions, i.e., the only inherent goal is to converse with the user on an unknown set of topics. While massive improvements in Natural Language Understanding (NLU) and the growth of available knowledge resources can partially support a robust conversation, these conversations generally lack the rapport between two humans that know each other. We developed a robust open-domain conversational system, Athena, that real Amazon Echo users access and evaluate at scale in the context of the Alexa Prize competition. We experiment with methods intended to increase intimacy between Athena and the user by heuristically developing a rule-based user model that personalizes both the current and subsequent conversations and evaluating specific personal opinion question strategies in A/B studies. Our results show a statistically significant positive impact on perceived conversation quality and length when employing these strategies.
comment: Won Best Paper at IWSDS '23
☆ Open World Classification with Adaptive Negative Samples EMNLP 2021
Open world classification is a task in natural language processing with key practical relevance and impact. Since the open or {\em unknown} category data only manifests in the inference phase, finding a model with a suitable decision boundary accommodating for the identification of known classes and discrimination of the open category is challenging. The performance of existing models is limited by the lack of effective open category data during the training stage or the lack of a good mechanism to learn appropriate decision boundaries. We propose an approach based on \underline{a}daptive \underline{n}egative \underline{s}amples (ANS) designed to generate effective synthetic open category samples in the training stage and without requiring any prior knowledge or external datasets. Empirically, we find a significant advantage in using auxiliary one-versus-rest binary classifiers, which effectively utilize the generated negative samples and avoid the complex threshold-seeking stage in previous works. Extensive experiments on three benchmark datasets show that ANS achieves significant improvements over state-of-the-art methods.
comment: Accepted by EMNLP 2021 (Main Track, Long Paper)
♻ ☆ Exploring the Feasibility of ChatGPT for Event Extraction
Event extraction is a fundamental task in natural language processing that involves identifying and extracting information about events mentioned in text. However, it is a challenging task due to the lack of annotated data, which is expensive and time-consuming to obtain. The emergence of large language models (LLMs) such as ChatGPT provides an opportunity to solve language tasks with simple prompts without the need for task-specific datasets and fine-tuning. While ChatGPT has demonstrated impressive results in tasks like machine translation, text summarization, and question answering, it presents challenges when used for complex tasks like event extraction. Unlike other tasks, event extraction requires the model to be provided with a complex set of instructions defining all event types and their schemas. To explore the feasibility of ChatGPT for event extraction and the challenges it poses, we conducted a series of experiments. Our results show that ChatGPT has, on average, only 51.04% of the performance of a task-specific model such as EEQA in long-tail and complex scenarios. Our usability testing experiments indicate that ChatGPT is not robust enough, and continuous refinement of the prompt does not lead to stable performance improvements, which can result in a poor user experience. Besides, ChatGPT is highly sensitive to different prompt styles.
♻ ☆ Complex QA and language models hybrid architectures, Survey
This paper reviews the state-of-the-art of hybrid language models architectures and strategies for "complex" question-answering (QA, CQA, CPS). Large Language Models (LLM) are good at leveraging public data on standard problems but once you want to tackle more specific complex questions or problems you may need specific architecture, knowledge, skills, methods, sensitive data protection, explainability, human approval and versatile feedback... We identify key elements augmenting LLM to solve complex questions or problems. We extend findings from the robust community edited research papers BIG, BLOOM and HELM which open source, benchmark and analyze limits and challenges of LLM in terms of tasks complexity and strict evaluation on accuracy (e.g. fairness, robustness, toxicity, ...). Recent projects like ChatGPT and GALACTICA have allowed non-specialists to grasp the great potential as well as the equally strong limitations of language models in complex QA. Hybridizing these models with different components could allow to overcome these different limits and go much further. We discuss some challenges associated with complex QA, including domain adaptation, decomposition and efficient multi-step QA, long form and non-factoid QA, safety and multi-sensitivity data protection, multimodal search, hallucinations, explainability and truthfulness, temproal reasoning. Therefore, we analyze current solutions and promising research trends, using elements such as: hybrid LLM architectures, active human reinforcement learning supervised with AI, prompting adaptation, neuro-symbolic and structured knowledge grounding, program synthesis, iterated decomposition and others.
♻ ☆ ELODIN: Naming Concepts in Embedding Spaces
Despite recent advancements, the field of text-to-image synthesis still suffers from lack of fine-grained control. Using only text, it remains challenging to deal with issues such as concept coherence and concept contamination. We propose a method to enhance control by generating specific concepts that can be reused throughout multiple images, effectively expanding natural language with new words that can be combined much like a painter's palette. Unlike previous contributions, our method does not copy visuals from input data and can generate concepts through text alone. We perform a set of comparisons that finds our method to be a significant improvement over text-only prompts.
comment: Added quantitative data, fixed formatting issues
♻ ☆ Linearly Mapping from Image to Text Space ICLR 2023
The extent to which text-only language models (LMs) learn to represent features of the non-linguistic world is an open question. Prior work has shown that pretrained LMs can be taught to caption images when a vision model's parameters are optimized to encode images in the language space. We test a stronger hypothesis: that the conceptual representations learned by frozen text-only models and vision-only models are similar enough that this can be achieved with a linear map. We show that the image representations from vision models can be transferred as continuous prompts to frozen LMs by training only a single linear projection. Using these to prompt the LM achieves competitive performance on captioning and visual question answering tasks compared to models that tune both the image encoder and text decoder (such as the MAGMA model). We compare three image encoders with increasing amounts of linguistic supervision seen during pretraining: BEIT (no linguistic information), NF-ResNET (lexical category information), and CLIP (full natural language descriptions). We find that all three encoders perform equally well at transferring visual property information to the language model (e.g., whether an animal is large or small), but that image encoders pretrained with linguistic supervision more saliently encode category information (e.g., distinguishing hippo vs. elephant) and thus perform significantly better on benchmark language-and-vision tasks. Our results indicate that LMs encode conceptual information structurally similarly to vision-based models, even those that are solely trained on images. Code is available here: https://github.com/jmerullo/limber
comment: Accepted at ICLR 2023
♻ ☆ MedKLIP: Medical Knowledge Enhanced Language-Image Pre-Training
In this paper, we consider enhancing medical visual-language pre-training (VLP) with domain-specific knowledge, by exploiting the paired image-text reports from the radiological daily practice. In particular, we make the following contributions: First, unlike existing works that directly process the raw reports, we adopt a novel triplet extraction module to extract the medical-related information, avoiding unnecessary complexity from language grammar and enhancing the supervision signals; Second, we propose a novel triplet encoding module with entity translation by querying a knowledge base, to exploit the rich domain knowledge in medical field, and implicitly build relationships between medical entities in the language embedding space; Third, we propose to use a Transformer-based fusion model for spatially aligning the entity description with visual signals at the image patch level, enabling the ability for medical diagnosis; Fourth, we conduct thorough experiments to validate the effectiveness of our architecture, and benchmark on numerous public benchmarks, e.g., ChestX-ray14, RSNA Pneumonia, SIIM-ACR Pneumothorax, COVIDx CXR-2, COVID Rural, and EdemaSeverity. In both zero-shot and fine-tuning settings, our model has demonstrated strong performance compared with the former methods on disease classification and grounding.
♻ ☆ Weight Averaging: A Simple Yet Effective Method to Overcome Catastrophic Forgetting in Automatic Speech Recognition ICASSP 2023
Adapting a trained Automatic Speech Recognition (ASR) model to new tasks results in catastrophic forgetting of old tasks, limiting the model's ability to learn continually and to be extended to new speakers, dialects, languages, etc. Focusing on End-to-End ASR, in this paper, we propose a simple yet effective method to overcome catastrophic forgetting: weight averaging. By simply taking the average of the previous and the adapted model, our method achieves high performance on both the old and new tasks. It can be further improved by introducing a knowledge distillation loss during the adaptation. We illustrate the effectiveness of our method on both monolingual and multilingual ASR. In both cases, our method strongly outperforms all baselines, even in its simplest form.
comment: Accepted at ICASSP 2023. 5 pages
♻ ☆ Continual Learning for Monolingual End-to-End Automatic Speech Recognition
Adapting Automatic Speech Recognition (ASR) models to new domains results in a deterioration of performance on the original domain(s), a phenomenon called Catastrophic Forgetting (CF). Even monolingual ASR models cannot be extended to new accents, dialects, topics, etc. without suffering from CF, making them unable to be continually enhanced without storing all past data. Fortunately, Continual Learning (CL) methods, which aim to enable continual adaptation while overcoming CF, can be used. In this paper, we implement an extensive number of CL methods for End-to-End ASR and test and compare their ability to extend a monolingual Hybrid CTC-Transformer model across four new tasks. We find that the best performing CL method closes the gap between the fine-tuned model (lower bound) and the model trained jointly on all tasks (upper bound) by more than 40%, while requiring access to only 0.6% of the original data.
comment: Published at EUSIPCO 2022. 5 pages, 1 figure
♻ ☆ Paraphrasing Techniques for Maritime QA system
There has been an increasing interest in incorporating Artificial Intelligence (AI) into Defence and military systems to complement and augment human intelligence and capabilities. However, much work still needs to be done toward achieving an effective human-machine partnership. This work is aimed at enhancing human-machine communications by developing a capability for automatically translating human natural language into a machine-understandable language (e.g., SQL queries). Techniques toward achieving this goal typically involve building a semantic parser trained on a very large amount of high-quality manually-annotated data. However, in many real-world Defence scenarios, it is not feasible to obtain such a large amount of training data. To the best of our knowledge, there are few works trying to explore the possibility of training a semantic parser with limited manually-paraphrased data, in other words, zero-shot. In this paper, we investigate how to exploit paraphrasing methods for the automated generation of large-scale training datasets (in the form of paraphrased utterances and their corresponding logical forms in SQL format) and present our experimental results using real-world data in the maritime domain.
comment: 8 pages. The first three authors contribute equally
♻ ☆ ViLPAct: A Benchmark for Compositional Generalization on Multimodal Human Activities EACL2023
We introduce ViLPAct, a novel vision-language benchmark for human activity planning. It is designed for a task where embodied AI agents can reason and forecast future actions of humans based on video clips about their initial activities and intents in text. The dataset consists of 2.9k videos from \charades extended with intents via crowdsourcing, a multi-choice question test set, and four strong baselines. One of the baselines implements a neurosymbolic approach based on a multi-modal knowledge base (MKB), while the other ones are deep generative models adapted from recent state-of-the-art (SOTA) methods. According to our extensive experiments, the key challenges are compositional generalization and effective use of information from both modalities.
comment: Accepted at EACL2023 (Findings)
♻ ☆ On Robustness of Prompt-based Semantic Parsing with Large Pre-trained Language Model: An Empirical Study on Codex EACL2023
Semantic parsing is a technique aimed at constructing a structured representation of the meaning of a natural-language question. Recent advancements in few-shot language models trained on code have demonstrated superior performance in generating these representations compared to traditional unimodal language models, which are trained on downstream tasks. Despite these advancements, existing fine-tuned neural semantic parsers are susceptible to adversarial attacks on natural-language inputs. While it has been established that the robustness of smaller semantic parsers can be enhanced through adversarial training, this approach is not feasible for large language models in real-world scenarios, as it requires both substantial computational resources and expensive human annotation on in-domain semantic parsing data. This paper presents the first empirical study on the adversarial robustness of a large prompt-based language model of code, \codex. Our results demonstrate that the state-of-the-art (SOTA) code-language models are vulnerable to carefully crafted adversarial examples. To address this challenge, we propose methods for improving robustness without the need for significant amounts of labeled data or heavy computational resources.
comment: Accepted at EACL2023 (main)
♻ ☆ Active Learning for Event Extraction with Memory-based Loss Prediction Model
Event extraction (EE) plays an important role in many industrial application scenarios, and high-quality EE methods require a large amount of manual annotation data to train supervised learning models. However, the cost of obtaining annotation data is very high, especially for annotation of domain events, which requires the participation of experts from corresponding domain. So we introduce active learning (AL) technology to reduce the cost of event annotation. But the existing AL methods have two main problems, which make them not well used for event extraction. Firstly, the existing pool-based selection strategies have limitations in terms of computational cost and sample validity. Secondly, the existing evaluation of sample importance lacks the use of local sample information. In this paper, we present a novel deep AL method for EE. We propose a batch-based selection strategy and a Memory-Based Loss Prediction model (MBLP) to select unlabeled samples efficiently. During the selection process, we use an internal-external sample loss ranking method to evaluate the sample importance by using local information. Finally, we propose a delayed training strategy to train the MBLP model. Extensive experiments are performed on three domain datasets, and our method outperforms other state-of-the-art methods.
♻ ☆ Challenges in Explanation Quality Evaluation
While much research focused on producing explanations, it is still unclear how the produced explanations' quality can be evaluated in a meaningful way. Today's predominant approach is to quantify explanations using proxy scores which compare explanations to (human-annotated) gold explanations. This approach assumes that explanations which reach higher proxy scores will also provide a greater benefit to human users. In this paper, we present problems of this approach. Concretely, we (i) formulate desired characteristics of explanation quality, (ii) describe how current evaluation practices violate them, and (iii) support our argumentation with initial evidence from a crowdsourcing case study in which we investigate the explanation quality of state-of-the-art explainable question answering systems. We find that proxy scores correlate poorly with human quality ratings and, additionally, become less expressive the more often they are used (i.e. following Goodhart's law). Finally, we propose guidelines to enable a meaningful evaluation of explanations to drive the development of systems that provide tangible benefits to human users.
comment: 41 pages, 11 figures
♻ ☆ A Challenging Benchmark for Low-Resource Learning
With promising yet saturated results in high-resource settings, low-resource datasets have gradually become popular benchmarks for evaluating the learning ability of advanced neural networks (e.g., BigBench, superGLUE). Some models even surpass humans according to benchmark test results. However, we find that there exists a set of hard examples in low-resource settings that challenge neural networks but are not well evaluated, which causes over-estimated performance. We first give a theoretical analysis on which factors bring the difficulty of low-resource learning. It then motivate us to propose a challenging benchmark hardBench to better evaluate the learning ability, which covers 11 datasets, including 3 computer vision (CV) datasets and 8 natural language process (NLP) datasets. Experiments on a wide range of models show that neural networks, even pre-trained language models, have sharp performance drops on our benchmark, demonstrating the effectiveness on evaluating the weaknesses of neural networks. On NLP tasks, we surprisingly find that despite better results on traditional low-resource benchmarks, pre-trained networks, does not show performance improvements on our benchmarks. These results demonstrate that there are still a large robustness gap between existing models and human-level performance.
♻ ☆ Flat Multi-modal Interaction Transformer for Named Entity Recognition COLING 2022
Multi-modal named entity recognition (MNER) aims at identifying entity spans and recognizing their categories in social media posts with the aid of images. However, in dominant MNER approaches, the interaction of different modalities is usually carried out through the alternation of self-attention and cross-attention or over-reliance on the gating machine, which results in imprecise and biased correspondence between fine-grained semantic units of text and image. To address this issue, we propose a Flat Multi-modal Interaction Transformer (FMIT) for MNER. Specifically, we first utilize noun phrases in sentences and general domain words to obtain visual cues. Then, we transform the fine-grained semantic representation of the vision and text into a unified lattice structure and design a novel relative position encoding to match different modalities in Transformer. Meanwhile, we propose to leverage entity boundary detection as an auxiliary task to alleviate visual bias. Experiments show that our methods achieve the new state-of-the-art performance on two benchmark datasets.
comment: Accepted by COLING 2022, oral paper
♻ ☆ On the Risks of Stealing the Decoding Algorithms of Language Models
A key component of generating text from modern language models (LM) is the selection and tuning of decoding algorithms. These algorithms determine how to generate text from the internal probability distribution generated by the LM. The process of choosing a decoding algorithm and tuning its hyperparameters takes significant time, manual effort, and computation, and it also requires extensive human evaluation. Therefore, the identity and hyperparameters of such decoding algorithms are considered to be extremely valuable to their owners. In this work, we show, for the first time, that an adversary with typical API access to an LM can steal the type and hyperparameters of its decoding algorithms at very low monetary costs. Our attack is effective against popular LMs used in text generation APIs, including GPT-2 and GPT-3. We demonstrate the feasibility of stealing such information with only a few dollars, e.g., $\$0.8$, $\$1$, $\$4$, and $\$40$ for the four versions of GPT-3.
♻ ☆ Mask-guided BERT for Few Shot Text Classification
Transformer-based language models have achieved significant success in various domains. However, the data-intensive nature of the transformer architecture requires much labeled data, which is challenging in low-resource scenarios (i.e., few-shot learning (FSL)). The main challenge of FSL is the difficulty of training robust models on small amounts of samples, which frequently leads to overfitting. Here we present Mask-BERT, a simple and modular framework to help BERT-based architectures tackle FSL. The proposed approach fundamentally differs from existing FSL strategies such as prompt tuning and meta-learning. The core idea is to selectively apply masks on text inputs and filter out irrelevant information, which guides the model to focus on discriminative tokens that influence prediction results. In addition, to make the text representations from different categories more separable and the text representations from the same category more compact, we introduce a contrastive learning loss function. Experimental results on public-domain benchmark datasets demonstrate the effectiveness of Mask-BERT.
Computer Vision and Pattern Recognition
☆ PAC-NeRF: Physics Augmented Continuum Neural Radiance Fields for Geometry-Agnostic System Identification ICLR 2023
Existing approaches to system identification (estimating the physical parameters of an object) from videos assume known object geometries. This precludes their applicability in a vast majority of scenes where object geometries are complex or unknown. In this work, we aim to identify parameters characterizing a physical system from a set of multi-view videos without any assumption on object geometry or topology. To this end, we propose "Physics Augmented Continuum Neural Radiance Fields" (PAC-NeRF), to estimate both the unknown geometry and physical parameters of highly dynamic objects from multi-view videos. We design PAC-NeRF to only ever produce physically plausible states by enforcing the neural radiance field to follow the conservation laws of continuum mechanics. For this, we design a hybrid Eulerian-Lagrangian representation of the neural radiance field, i.e., we use the Eulerian grid representation for NeRF density and color fields, while advecting the neural radiance fields via Lagrangian particles. This hybrid Eulerian-Lagrangian representation seamlessly blends efficient neural rendering with the material point method (MPM) for robust differentiable physics simulation. We validate the effectiveness of our proposed framework on geometry and physical parameter estimation over a vast range of materials, including elastic bodies, plasticine, sand, Newtonian and non-Newtonian fluids, and demonstrate significant performance gain on most tasks.
comment: ICLR 2023 Spotlight. Project page: https://sites.google.com/view/PAC-NeRF
☆ Scaling up GANs for Text-to-Image Synthesis CVPR 2023
The recent success of text-to-image synthesis has taken the world by storm and captured the general public's imagination. From a technical standpoint, it also marked a drastic change in the favored architecture to design generative image models. GANs used to be the de facto choice, with techniques like StyleGAN. With DALL-E 2, auto-regressive and diffusion models became the new standard for large-scale generative models overnight. This rapid shift raises a fundamental question: can we scale up GANs to benefit from large datasets like LAION? We find that na\"Ively increasing the capacity of the StyleGAN architecture quickly becomes unstable. We introduce GigaGAN, a new GAN architecture that far exceeds this limit, demonstrating GANs as a viable option for text-to-image synthesis. GigaGAN offers three major advantages. First, it is orders of magnitude faster at inference time, taking only 0.13 seconds to synthesize a 512px image. Second, it can synthesize high-resolution images, for example, 16-megapixel pixels in 3.66 seconds. Finally, GigaGAN supports various latent space editing applications such as latent interpolation, style mixing, and vector arithmetic operations.
comment: CVPR 2023. Project webpage at https://mingukkang.github.io/GigaGAN/
☆ Open-world Instance Segmentation: Top-down Learning with Bottom-up Supervision
Many top-down architectures for instance segmentation achieve significant success when trained and tested on pre-defined closed-world taxonomy. However, when deployed in the open world, they exhibit notable bias towards seen classes and suffer from significant performance drop. In this work, we propose a novel approach for open world instance segmentation called bottom-Up and top-Down Open-world Segmentation (UDOS) that combines classical bottom-up segmentation algorithms within a top-down learning framework. UDOS first predicts parts of objects using a top-down network trained with weak supervision from bottom-up segmentations. The bottom-up segmentations are class-agnostic and do not overfit to specific taxonomies. The part-masks are then fed into affinity-based grouping and refinement modules to predict robust instance-level segmentations. UDOS enjoys both the speed and efficiency from the top-down architectures and the generalization ability to unseen categories from bottom-up supervision. We validate the strengths of UDOS on multiple cross-category as well as cross-dataset transfer tasks from 5 challenging datasets including MS-COCO, LVIS, ADE20k, UVO and OpenImages, achieving significant improvements over state-of-the-art across the board. Our code and models are available on our project page.
comment: Project page: https://tarun005.github.io/UDOS
☆ Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detection
In this paper, we present an open-set object detector, called Grounding DINO, by marrying Transformer-based detector DINO with grounded pre-training, which can detect arbitrary objects with human inputs such as category names or referring expressions. The key solution of open-set object detection is introducing language to a closed-set detector for open-set concept generalization. To effectively fuse language and vision modalities, we conceptually divide a closed-set detector into three phases and propose a tight fusion solution, which includes a feature enhancer, a language-guided query selection, and a cross-modality decoder for cross-modality fusion. While previous works mainly evaluate open-set object detection on novel categories, we propose to also perform evaluations on referring expression comprehension for objects specified with attributes. Grounding DINO performs remarkably well on all three settings, including benchmarks on COCO, LVIS, ODinW, and RefCOCO/+/g. Grounding DINO achieves a $52.5$ AP on the COCO detection zero-shot transfer benchmark, i.e., without any training data from COCO. It sets a new record on the ODinW zero-shot benchmark with a mean $26.1$ AP. Code will be available at \url{https://github.com/IDEA-Research/GroundingDINO}.
☆ Mark My Words: Dangers of Watermarked Images in ImageNet ICLR 2023
The utilization of pre-trained networks, especially those trained on ImageNet, has become a common practice in Computer Vision. However, prior research has indicated that a significant number of images in the ImageNet dataset contain watermarks, making pre-trained networks susceptible to learning artifacts such as watermark patterns within their latent spaces. In this paper, we aim to assess the extent to which popular pre-trained architectures display such behavior and to determine which classes are most affected. Additionally, we examine the impact of watermarks on the extracted features. Contrary to the popular belief that the Chinese logographic watermarks impact the "carton" class only, our analysis reveals that a variety of ImageNet classes, such as "monitor", "broom", "apron" and "safe" rely on spurious correlations. Finally, we propose a simple approach to mitigate this issue in fine-tuned networks by ignoring the encodings from the feature-extractor layer of ImageNet pre-trained networks that are most susceptible to watermark imprints.
comment: 5 pages, 4 figures, Accepted to the ICLR 2023 TrustML-(un)Limited workshop
☆ Mimic before Reconstruct: Enhancing Masked Autoencoders with Feature Mimicking
Masked Autoencoders (MAE) have been popular paradigms for large-scale vision representation pre-training. However, MAE solely reconstructs the low-level RGB signals after the decoder and lacks supervision upon high-level semantics for the encoder, thus suffering from sub-optimal learned representations and long pre-training epochs. To alleviate this, previous methods simply replace the pixel reconstruction targets of 75% masked tokens by encoded features from pre-trained image-image (DINO) or image-language (CLIP) contrastive learning. Different from those efforts, we propose to Mimic before Reconstruct for Masked Autoencoders, named as MR-MAE, which jointly learns high-level and low-level representations without interference during pre-training. For high-level semantics, MR-MAE employs a mimic loss over 25% visible tokens from the encoder to capture the pre-trained patterns encoded in CLIP and DINO. For low-level structures, we inherit the reconstruction loss in MAE to predict RGB pixel values for 75% masked tokens after the decoder. As MR-MAE applies high-level and low-level targets respectively at different partitions, the learning conflicts between them can be naturally overcome and contribute to superior visual representations for various downstream tasks. On ImageNet-1K, the MR-MAE base pre-trained for only 400 epochs achieves 85.8% top-1 accuracy after fine-tuning, surpassing the 1600-epoch MAE base by +2.2% and the previous state-of-the-art BEiT V2 base by +0.3%. Code and pre-trained models will be released at https://github.com/Alpha-VL/ConvMAE.
comment: 12 pages, 3 figures
☆ Spawrious: A Benchmark for Fine Control of Spurious Correlation Biases
The problem of spurious correlations (SCs) arises when a classifier relies on non-predictive features that happen to be correlated with the labels in the training data. For example, a classifier may misclassify dog breeds based on the background of dog images. This happens when the backgrounds are correlated with other breeds in the training data, leading to misclassifications during test time. Previous SC benchmark datasets suffer from varying issues, e.g., over-saturation or only containing one-to-one (O2O) SCs, but no many-to-many (M2M) SCs arising between groups of spurious attributes and classes. In this paper, we present Spawrious-{O2O, M2M}-{Easy, Medium, Hard}, an image classification benchmark suite containing spurious correlations among different dog breeds and background locations. To create this dataset, we employ a text-to-image model to generate photo-realistic images, and an image captioning model to filter out unsuitable ones. The resulting dataset is of high quality, containing approximately 152,000 images. Our experimental results demonstrate that state-of-the-art group robustness methods struggle with Spawrious, most notably on the Hard-splits with $<60\%$ accuracy. By examining model misclassifications, we detect reliances on spurious backgrounds, demonstrating that our dataset provides a significant challenge to drive future research.
☆ Understanding the Challenges and Opportunities of Pose-based Anomaly Detection
Pose-based anomaly detection is a video-analysis technique for detecting anomalous events or behaviors by examining human pose extracted from the video frames. Utilizing pose data alleviates privacy and ethical issues. Also, computation-wise, the complexity of pose-based models is lower than pixel-based approaches. However, it introduces more challenges, such as noisy skeleton data, losing important pixel information, and not having enriched enough features. These problems are exacerbated by a lack of anomaly detection datasets that are good enough representatives of real-world scenarios. In this work, we analyze and quantify the characteristics of two well-known video anomaly datasets to better understand the difficulties of pose-based anomaly detection. We take a step forward, exploring the discriminating power of pose and trajectory for video anomaly detection and their effectiveness based on context. We believe these experiments are beneficial for a better comprehension of pose-based anomaly detection and the datasets currently available. This will aid researchers in tackling the task of anomaly detection with a more lucid perspective, accelerating the development of robust models with better performance.
☆ Presentation Attack Detection with Advanced CNN Models for Noncontact-based Fingerprint Systems
Touch-based fingerprint biometrics is one of the most popular biometric modalities with applications in several fields. Problems associated with touch-based techniques such as the presence of latent fingerprints and hygiene issues due to many people touching the same surface motivated the community to look for non-contact-based solutions. For the last few years, contactless fingerprint systems are on the rise and in demand because of the ability to turn any device with a camera into a fingerprint reader. Yet, before we can fully utilize the benefit of noncontact-based methods, the biometric community needs to resolve a few concerns such as the resiliency of the system against presentation attacks. One of the major obstacles is the limited publicly available data sets with inadequate spoof and live data. In this publication, we have developed a Presentation attack detection (PAD) dataset of more than 7500 four-finger images and more than 14,000 manually segmented single-fingertip images, and 10,000 synthetic fingertips (deepfakes). The PAD dataset was collected from six different Presentation Attack Instruments (PAI) of three different difficulty levels according to FIDO protocols, with five different types of PAI materials, and different smartphone cameras with manual focusing. We have utilized DenseNet-121 and NasNetMobile models and our proposed dataset to develop PAD algorithms and achieved PAD accuracy of Attack presentation classification error rate (APCER) 0.14\% and Bonafide presentation classification error rate (BPCER) 0.18\%. We have also reported the test results of the models against unseen spoof types to replicate uncertain real-world testing scenarios.
☆ Kernel Regression with Infinite-Width Neural Networks on Millions of Examples
Neural kernels have drastically increased performance on diverse and nonstandard data modalities but require significantly more compute, which previously limited their application to smaller datasets. In this work, we address this by massively parallelizing their computation across many GPUs. We combine this with a distributed, preconditioned conjugate gradients algorithm to enable kernel regression at a large scale (i.e. up to five million examples). Using this approach, we study scaling laws of several neural kernels across many orders of magnitude for the CIFAR-5m dataset. Using data augmentation to expand the original CIFAR-10 training dataset by a factor of 20, we obtain a test accuracy of 91.2\% (SotA for a pure kernel method). Moreover, we explore neural kernels on other data modalities, obtaining results on protein and small molecule prediction tasks that are competitive with SotA methods.
☆ FaceXHuBERT: Text-less Speech-driven E(X)pressive 3D Facial Animation Synthesis Using Self-Supervised Speech Representation Learning
This paper presents FaceXHuBERT, a text-less speech-driven 3D facial animation generation method that allows to capture personalized and subtle cues in speech (e.g. identity, emotion and hesitation). It is also very robust to background noise and can handle audio recorded in a variety of situations (e.g. multiple people speaking). Recent approaches employ end-to-end deep learning taking into account both audio and text as input to generate facial animation for the whole face. However, scarcity of publicly available expressive audio-3D facial animation datasets poses a major bottleneck. The resulting animations still have issues regarding accurate lip-synching, expressivity, person-specific information and generalizability. We effectively employ self-supervised pretrained HuBERT model in the training process that allows us to incorporate both lexical and non-lexical information in the audio without using a large lexicon. Additionally, guiding the training with a binary emotion condition and speaker identity distinguishes the tiniest subtle facial motion. We carried out extensive objective and subjective evaluation in comparison to ground-truth and state-of-the-art work. A perceptual user study demonstrates that our approach produces superior results with respect to the realism of the animation 78% of the time in comparison to the state-of-the-art. In addition, our method is 4 times faster eliminating the use of complex sequential models such as transformers. We strongly recommend watching the supplementary video before reading the paper. We also provide the implementation and evaluation codes with a GitHub repository link.
comment: 13 pages, 4 figures, code included
☆ 3DGen: Triplane Latent Diffusion for Textured Mesh Generation
Latent diffusion models for image generation have crossed a quality threshold which enabled them to achieve mass adoption. Recently, a series of works have made advancements towards replicating this success in the 3D domain, introducing techniques such as point cloud VAE, triplane representation, neural implicit surfaces and differentiable rendering based training. We take another step along this direction, combining these developments in a two-step pipeline consisting of 1) a triplane VAE which can learn latent representations of textured meshes and 2) a conditional diffusion model which generates the triplane features. For the first time this architecture allows conditional and unconditional generation of high quality textured or untextured 3D meshes across multiple diverse categories in a few seconds on a single GPU. It outperforms previous work substantially on image-conditioned and unconditional generation on mesh quality as well as texture generation. Furthermore, we demonstrate the scalability of our model to large datasets for increased quality and diversity. We will release our code and trained models.
☆ Rethinking Self-Supervised Visual Representation Learning in Pre-training for 3D Human Pose and Shape Estimation ICLR 2023
Recently, a few self-supervised representation learning (SSL) methods have outperformed the ImageNet classification pre-training for vision tasks such as object detection. However, its effects on 3D human body pose and shape estimation (3DHPSE) are open to question, whose target is fixed to a unique class, the human, and has an inherent task gap with SSL. We empirically study and analyze the effects of SSL and further compare it with other pre-training alternatives for 3DHPSE. The alternatives are 2D annotation-based pre-training and synthetic data pre-training, which share the motivation of SSL that aims to reduce the labeling cost. They have been widely utilized as a source of weak-supervision or fine-tuning, but have not been remarked as a pre-training source. SSL methods underperform the conventional ImageNet classification pre-training on multiple 3DHPSE benchmarks by 7.7% on average. In contrast, despite a much less amount of pre-training data, the 2D annotation-based pre-training improves accuracy on all benchmarks and shows faster convergence during fine-tuning. Our observations challenge the naive application of the current SSL pre-training to 3DHPSE and relight the value of other data types in the pre-training aspect.
comment: Accepted to ICLR 2023, 18 pages including the appendix
☆ Rethinking Range View Representation for LiDAR Segmentation
LiDAR segmentation is crucial for autonomous driving perception. Recent trends favor point- or voxel-based methods as they often yield better performance than the traditional range view representation. In this work, we unveil several key factors in building powerful range view models. We observe that the "many-to-one" mapping, semantic incoherence, and shape deformation are possible impediments against effective learning from range view projections. We present RangeFormer -- a full-cycle framework comprising novel designs across network architecture, data augmentation, and post-processing -- that better handles the learning and processing of LiDAR point clouds from the range view. We further introduce a Scalable Training from Range view (STR) strategy that trains on arbitrary low-resolution 2D range images, while still maintaining satisfactory 3D segmentation accuracy. We show that, for the first time, a range view method is able to surpass the point, voxel, and multi-view fusion counterparts in the competing LiDAR semantic and panoptic segmentation benchmarks, i.e., SemanticKITTI, nuScenes, and ScribbleKITTI.
comment: 19 pages, 9 figures, 11 tables, project website at https://ldkong.com/RangeFormer
☆ Knowledge-augmented Few-shot Visual Relation Detection
Visual Relation Detection (VRD) aims to detect relationships between objects for image understanding. Most existing VRD methods rely on thousands of training samples of each relationship to achieve satisfactory performance. Some recent papers tackle this problem by few-shot learning with elaborately designed pipelines and pre-trained word vectors. However, the performance of existing few-shot VRD models is severely hampered by the poor generalization capability, as they struggle to handle the vast semantic diversity of visual relationships. Nonetheless, humans have the ability to learn new relationships with just few examples based on their knowledge. Inspired by this, we devise a knowledge-augmented, few-shot VRD framework leveraging both textual knowledge and visual relation knowledge to improve the generalization ability of few-shot VRD. The textual knowledge and visual relation knowledge are acquired from a pre-trained language model and an automatically constructed visual relation knowledge graph, respectively. We extensively validate the effectiveness of our framework. Experiments conducted on three benchmarks from the commonly used Visual Genome dataset show that our performance surpasses existing state-of-the-art models with a large improvement.
comment: work in progress
☆ Brain-Diffuser: Natural scene reconstruction from fMRI signals using generative latent diffusion
In neural decoding research, one of the most intriguing topics is the reconstruction of perceived natural images based on fMRI signals. Previous studies have succeeded in re-creating different aspects of the visuals, such as low-level properties (shape, texture, layout) or high-level features (category of objects, descriptive semantics of scenes) but have typically failed to reconstruct these properties together for complex scene images. Generative AI has recently made a leap forward with latent diffusion models capable of generating high-complexity images. Here, we investigate how to take advantage of this innovative technology for brain decoding. We present a two-stage scene reconstruction framework called ``Brain-Diffuser''. In the first stage, starting from fMRI signals, we reconstruct images that capture low-level properties and overall layout using a VDVAE (Very Deep Variational Autoencoder) model. In the second stage, we use the image-to-image framework of a latent diffusion model (Versatile Diffusion) conditioned on predicted multimodal (text and visual) features, to generate final reconstructed images. On the publicly available Natural Scenes Dataset benchmark, our method outperforms previous models both qualitatively and quantitatively. When applied to synthetic fMRI patterns generated from individual ROI (region-of-interest) masks, our trained model creates compelling ``ROI-optimal'' scenes consistent with neuroscientific knowledge. Thus, the proposed methodology can have an impact on both applied (e.g. brain-computer interface) and fundamental neuroscience.
☆ Adaptive Calibrator Ensemble for Model Calibration under Distribution Shift
Model calibration usually requires optimizing some parameters (e.g., temperature) w.r.t an objective function (e.g., negative log-likelihood). In this paper, we report a plain, important but often neglected fact that the objective function is influenced by calibration set difficulty, i.e., the ratio of the number of incorrectly classified samples to that of correctly classified samples. If a test set has a drastically different difficulty level from the calibration set, the optimal calibration parameters of the two datasets would be different. In other words, a calibrator optimal on the calibration set would be suboptimal on the OOD test set and thus has degraded performance. With this knowledge, we propose a simple and effective method named adaptive calibrator ensemble (ACE) to calibrate OOD datasets whose difficulty is usually higher than the calibration set. Specifically, two calibration functions are trained, one for in-distribution data (low difficulty), and the other for severely OOD data (high difficulty). To achieve desirable calibration on a new OOD dataset, ACE uses an adaptive weighting method that strikes a balance between the two extreme functions. When plugged in, ACE generally improves the performance of a few state-of-the-art calibration schemes on a series of OOD benchmarks. Importantly, such improvement does not come at the cost of the in-distribution calibration accuracy.
comment: 16 pages, 9 figures
☆ Tucker Bilinear Attention Network for Multi-scale Remote Sensing Object Detection
Object detection on VHR remote sensing images plays a vital role in applications such as urban planning, land resource management, and rescue missions. The large-scale variation of the remote-sensing targets is one of the main challenges in VHR remote-sensing object detection. Existing methods improve the detection accuracy of high-resolution remote sensing objects by improving the structure of feature pyramids and adopting different attention modules. However, for small targets, there still be seriously missed detections due to the loss of key detail features. There is still room for improvement in the way of multiscale feature fusion and balance. To address this issue, this paper proposes two novel modules: Guided Attention and Tucker Bilinear Attention, which are applied to the stages of early fusion and late fusion respectively. The former can effectively retain clean key detail features, and the latter can better balance features through semantic-level correlation mining. Based on two modules, we build a new multi-scale remote sensing object detection framework. No bells and whistles. The proposed method largely improves the average precisions of small objects and achieves the highest mean average precisions compared with 9 state-of-the-art methods on DOTA, DIOR, and NWPU VHR-10.Code and models are available at https://github.com/Shinichict/GTNet.
comment: arXiv admin note: text overlap with arXiv:1705.06676, arXiv:2209.13351 by other authors
☆ BaDLAD: A Large Multi-Domain Bengali Document Layout Analysis Dataset
While strides have been made in deep learning based Bengali Optical Character Recognition (OCR) in the past decade, the absence of large Document Layout Analysis (DLA) datasets has hindered the application of OCR in document transcription, e.g., transcribing historical documents and newspapers. Moreover, rule-based DLA systems that are currently being employed in practice are not robust to domain variations and out-of-distribution layouts. To this end, we present the first multidomain large Bengali Document Layout Analysis Dataset: BaDLAD. This dataset contains 33,695 human annotated document samples from six domains - i) books and magazines, ii) public domain govt. documents, iii) liberation war documents, iv) newspapers, v) historical newspapers, and vi) property deeds, with 710K polygon annotations for four unit types: text-box, paragraph, image, and table. Through preliminary experiments benchmarking the performance of existing state-of-the-art deep learning architectures for English DLA, we demonstrate the efficacy of our dataset in training deep learning based Bengali document digitization models.
☆ Controllable Video Generation by Learning the Underlying Dynamical System with Neural ODE
Videos depict the change of complex dynamical systems over time in the form of discrete image sequences. Generating controllable videos by learning the dynamical system is an important yet underexplored topic in the computer vision community. This paper presents a novel framework, TiV-ODE, to generate highly controllable videos from a static image and a text caption. Specifically, our framework leverages the ability of Neural Ordinary Differential Equations~(Neural ODEs) to represent complex dynamical systems as a set of nonlinear ordinary differential equations. The resulting framework is capable of generating videos with both desired dynamics and content. Experiments demonstrate the ability of the proposed method in generating highly controllable and visually consistent videos, and its capability of modeling dynamical systems. Overall, this work is a significant step towards developing advanced controllable video generation models that can handle complex and dynamic scenes.
☆ WASD: A Wilder Active Speaker Detection Dataset
Current Active Speaker Detection (ASD) models achieve great results on AVA-ActiveSpeaker (AVA), using only sound and facial features. Although this approach is applicable in movie setups (AVA), it is not suited for less constrained conditions. To demonstrate this limitation, we propose a Wilder Active Speaker Detection (WASD) dataset, with increased difficulty by targeting the two key components of current ASD: audio and face. Grouped into 5 categories, ranging from optimal conditions to surveillance settings, WASD contains incremental challenges for ASD with tactical impairment of audio and face data. We select state-of-the-art models and assess their performance in two groups of WASD: Easy (cooperative settings) and Hard (audio and/or face are specifically degraded). The results show that: 1) AVA trained models maintain a state-of-the-art performance in WASD Easy group, while underperforming in the Hard one, showing the 2) similarity between AVA and Easy data; and 3) training in WASD does not improve models performance to AVA levels, particularly for audio impairment and surveillance settings. This shows that AVA does not prepare models for wild ASD and current approaches are subpar to deal with such conditions. The proposed dataset also contains body data annotations to provide a new source for ASD, and is available at https://github.com/Tiago-Roxo/WASD.
☆ Replacement as a Self-supervision for Fine-grained Vision-language Pre-training
Fine-grained supervision based on object annotations has been widely used for vision and language pre-training (VLP). However, in real-world application scenarios, aligned multi-modal data is usually in the image-caption format, which only provides coarse-grained supervision. It is cost-expensive to collect object annotations and build object annotation pre-extractor for different scenarios. In this paper, we propose a fine-grained self-supervision signal without object annotations from a replacement perspective. First, we propose a homonym sentence rewriting (HSR) algorithm to provide token-level supervision. The algorithm replaces a verb/noun/adjective/quantifier word of the caption with its homonyms from WordNet. Correspondingly, we propose a replacement vision-language modeling (RVLM) framework to exploit the token-level supervision. Two replaced modeling tasks, i.e., replaced language contrastive (RLC) and replaced language modeling (RLM), are proposed to learn the fine-grained alignment. Extensive experiments on several downstream tasks demonstrate the superior performance of the proposed method.
comment: Work in progress
☆ 3D Video Loops from Asynchronous Input
Looping videos are short video clips that can be looped endlessly without visible seams or artifacts. They provide a very attractive way to capture the dynamism of natural scenes. Existing methods have been mostly limited to 2D representations. In this paper, we take a step forward and propose a practical solution that enables an immersive experience on dynamic 3D looping scenes. The key challenge is to consider the per-view looping conditions from asynchronous input while maintaining view consistency for the 3D representation. We propose a novel sparse 3D video representation, namely Multi-Tile Video (MTV), which not only provides a view-consistent prior, but also greatly reduces memory usage, making the optimization of a 4D volume tractable. Then, we introduce a two-stage pipeline to construct the 3D looping MTV from completely asynchronous multi-view videos with no time overlap. A novel looping loss based on video temporal retargeting algorithms is adopted during the optimization to loop the 3D scene. Experiments of our framework have shown promise in successfully generating and rendering photorealistic 3D looping videos in real time even on mobile devices. The code, dataset, and live demos are available in https://limacv.github.io/VideoLoop3D_web/.
comment: For more information, please visit the homepage at https://limacv.github.io/VideoLoop3D_web/
☆ MixSpeech: Cross-Modality Self-Learning with Audio-Visual Stream Mixup for Visual Speech Translation and Recognition
Multi-media communications facilitate global interaction among people. However, despite researchers exploring cross-lingual translation techniques such as machine translation and audio speech translation to overcome language barriers, there is still a shortage of cross-lingual studies on visual speech. This lack of research is mainly due to the absence of datasets containing visual speech and translated text pairs. In this paper, we present \textbf{AVMuST-TED}, the first dataset for \textbf{A}udio-\textbf{V}isual \textbf{Mu}ltilingual \textbf{S}peech \textbf{T}ranslation, derived from \textbf{TED} talks. Nonetheless, visual speech is not as distinguishable as audio speech, making it difficult to develop a mapping from source speech phonemes to the target language text. To address this issue, we propose MixSpeech, a cross-modality self-learning framework that utilizes audio speech to regularize the training of visual speech tasks. To further minimize the cross-modality gap and its impact on knowledge transfer, we suggest adopting mixed speech, which is created by interpolating audio and visual streams, along with a curriculum learning strategy to adjust the mixing ratio as needed. MixSpeech enhances speech translation in noisy environments, improving BLEU scores for four languages on AVMuST-TED by +1.4 to +4.2. Moreover, it achieves state-of-the-art performance in lip reading on CMLR (11.1\%), LRS2 (25.5\%), and LRS3 (28.0\%).
comment: https://github.com/Exgc/AVMuST-TED
☆ SpyroPose: Importance Sampling Pyramids for Object Pose Distribution Estimation in SE(3)
Object pose estimation is a core computer vision problem and often an essential component in robotics. Pose estimation is usually approached by seeking the single best estimate of an object's pose, but this approach is ill-suited for tasks involving visual ambiguity. In such cases it is desirable to estimate the uncertainty as a pose distribution to allow downstream tasks to make informed decisions. Pose distributions can have arbitrary complexity which motivates estimating unparameterized distributions, however, until now they have only been used for orientation estimation on SO(3) due to the difficulty in training on and normalizing over SE(3). We propose a novel method for pose distribution estimation on SE(3). We use a hierarchical grid, a pyramid, which enables efficient importance sampling during training and sparse evaluation of the pyramid at inference, allowing real time 6D pose distribution estimation. Our method outperforms state-of-the-art methods on SO(3), and to the best of our knowledge, we provide the first quantitative results on pose distribution estimation on SE(3). Code will be available at spyropose.github.io
comment: 8 pages
☆ National-scale 1-m resolution land-cover mapping for the entire China based on a low-cost solution and open-access data
Nowadays, many large-scale land-cover (LC) products have been released, however, current LC products for China either lack a fine resolution or nationwide coverage. With the rapid urbanization of China, there is an urgent need for creating a very-high-resolution (VHR) national-scale LC map for China. In this study, a novel 1-m resolution LC map of China covering $9,600,000 km^2$, called SinoLC-1, was produced by using a deep learning framework and multi-source open-access data. To efficiently generate the VHR national-scale LC map, firstly, the reliable LC labels were collected from three 10-m LC products and Open Street Map data. Secondly, the collected 10-m labels and 1-m Google Earth imagery were utilized in the proposed low-to-high (L2H) framework for training. With weak and self-supervised strategies, the L2H framework resolves the label noise brought by the mismatched resolution between training pairs and produces VHR results. Lastly, we compare the SinoLC-1 with five widely used products and validate it with a sample set including 10,6852 points and a statistical report collected from the government. The results show the SinoLC-1 achieved an OA of 74\% and a Kappa of 0.65. Moreover, as the first 1-m national-scale LC map for China, the SinoLC-1 shows overall acceptable results with the finest landscape details.
comment: 4 pages, 3 figures, conference paper
☆ M3AE: Multimodal Representation Learning for Brain Tumor Segmentation with Missing Modalities
Multimodal magnetic resonance imaging (MRI) provides complementary information for sub-region analysis of brain tumors. Plenty of methods have been proposed for automatic brain tumor segmentation using four common MRI modalities and achieved remarkable performance. In practice, however, it is common to have one or more modalities missing due to image corruption, artifacts, acquisition protocols, allergy to contrast agents, or simply cost. In this work, we propose a novel two-stage framework for brain tumor segmentation with missing modalities. In the first stage, a multimodal masked autoencoder (M3AE) is proposed, where both random modalities (i.e., modality dropout) and random patches of the remaining modalities are masked for a reconstruction task, for self-supervised learning of robust multimodal representations against missing modalities. To this end, we name our framework M3AE. Meanwhile, we employ model inversion to optimize a representative full-modal image at marginal extra cost, which will be used to substitute for the missing modalities and boost performance during inference. Then in the second stage, a memory-efficient self distillation is proposed to distill knowledge between heterogenous missing-modal situations while fine-tuning the model for supervised segmentation. Our M3AE belongs to the 'catch-all' genre where a single model can be applied to all possible subsets of modalities, thus is economic for both training and deployment. Extensive experiments on BraTS 2018 and 2020 datasets demonstrate its superior performance to existing state-of-the-art methods with missing modalities, as well as the efficacy of its components. Our code is available at: https://github.com/ccarliu/m3ae.
☆ Perspective Projection-Based 3D CT Reconstruction from Biplanar X-rays
X-ray computed tomography (CT) is one of the most common imaging techniques used to diagnose various diseases in the medical field. Its high contrast sensitivity and spatial resolution allow the physician to observe details of body parts such as bones, soft tissue, blood vessels, etc. As it involves potentially harmful radiation exposure to patients and surgeons, however, reconstructing 3D CT volume from perpendicular 2D X-ray images is considered a promising alternative, thanks to its lower radiation risk and better accessibility. This is highly challenging though, since it requires reconstruction of 3D anatomical information from 2D images with limited views, where all the information is overlapped. In this paper, we propose PerX2CT, a novel CT reconstruction framework from X-ray that reflects the perspective projection scheme. Our proposed method provides a different combination of features for each coordinate which implicitly allows the model to obtain information about the 3D location. We reveal the potential to reconstruct the selected part of CT with high resolution by properly using the coordinate-wise local and global features. Our approach shows potential for use in clinical applications with low computational complexity and fast inference time, demonstrating superior performance than baselines in multiple evaluation metrics.
☆ Detecting Images Generated by Diffusers
This paper explores the task of detecting images generated by text-to-image diffusion models. To evaluate this, we consider images generated from captions in the MSCOCO and Wikimedia datasets using two state-of-the-art models: Stable Diffusion and GLIDE. Our experiments show that it is possible to detect the generated images using simple Multi-Layer Perceptrons (MLPs), starting from features extracted by CLIP, or traditional Convolutional Neural Networks (CNNs). We also observe that models trained on images generated by Stable Diffusion can detect images generated by GLIDE relatively well, however, the reverse is not true. Lastly, we find that incorporating the associated textual information with the images rarely leads to significant improvement in detection results but that the type of subject depicted in the image can have a significant impact on performance. This work provides insights into the feasibility of detecting generated images, and has implications for security and privacy concerns in real-world applications.
☆ Effective Pseudo-Labeling based on Heatmap for Unsupervised Domain Adaptation in Cell Detection
Cell detection is an important task in biomedical research. Recently, deep learning methods have made it possible to improve the performance of cell detection. However, a detection network trained with training data under a specific condition (source domain) may not work well on data under other conditions (target domains), which is called the domain shift problem. In particular, cells are cultured under different conditions depending on the purpose of the research. Characteristics, e.g., the shapes and density of the cells, change depending on the conditions, and such changes may cause domain shift problems. Here, we propose an unsupervised domain adaptation method for cell detection using a pseudo-cell-position heatmap, where the cell centroid is at the peak of a Gaussian distribution in the map and selective pseudo-labeling. In the prediction result for the target domain, even if the peak location is correct, the signal distribution around the peak often has a non-Gaussian shape. The pseudo-cell-position heatmap is thus re-generated using the peak positions in the predicted heatmap to have a clear Gaussian shape. Our method selects confident pseudo-cell-position heatmaps based on uncertainty and curriculum learning. We conducted numerous experiments showing that, compared with the existing methods, our method improved detection performance under different conditions.
comment: 16 pages, 18 figures, Accepted in Medical Image Analysis 2022
☆ From Visual Prompt Learning to Zero-Shot Transfer: Mapping Is All You Need
Visual prompt learning, as a newly emerged technique, leverages the knowledge learned by a large-scale pre-trained model and adapts it to downstream tasks through the usage of prompts. While previous research has focused on designing effective prompts, in this work, we argue that compared to prompt design, a good mapping strategy matters more. In this sense, we propose SeMap, a more effective mapping using the semantic alignment between the pre-trained model's knowledge and the downstream task. Our experimental results show that SeMap can largely boost the performance of visual prompt learning. Moreover, our experiments show that SeMap is capable of achieving competitive zero-shot transfer, indicating that it can perform the downstream task without any fine-tuning on the corresponding dataset. This demonstrates the potential of our proposed method to be used in a broader range of applications where the zero-shot transfer is desired. Results suggest that our proposed SeMap could lead to significant advancements in both visual prompt learning and zero-shot transfer. We hope with SeMap, we can help the community move forward to more efficient and lightweight utilization of large vision models.
☆ Masked Image Modeling with Local Multi-Scale Reconstruction CVPR 2023
Masked Image Modeling (MIM) achieves outstanding success in self-supervised representation learning. Unfortunately, MIM models typically have huge computational burden and slow learning process, which is an inevitable obstacle for their industrial applications. Although the lower layers play the key role in MIM, existing MIM models conduct reconstruction task only at the top layer of encoder. The lower layers are not explicitly guided and the interaction among their patches is only used for calculating new activations. Considering the reconstruction task requires non-trivial inter-patch interactions to reason target signals, we apply it to multiple local layers including lower and upper layers. Further, since the multiple layers expect to learn the information of different scales, we design local multi-scale reconstruction, where the lower and upper layers reconstruct fine-scale and coarse-scale supervision signals respectively. This design not only accelerates the representation learning process by explicitly guiding multiple layers, but also facilitates multi-scale semantical understanding to the input. Extensive experiments show that with significantly less pre-training burden, our model achieves comparable or better performance on classification, detection and segmentation tasks than existing MIM models.
comment: CVPR 2023
☆ Probabilistic 3d regression with projected huber distribution
Estimating probability distributions which describe where an object is likely to be from camera data is a task with many applications. In this work we describe properties which we argue such methods should conform to. We also design a method which conform to these properties. In our experiments we show that our method produces uncertainties which correlate well with empirical errors. We also show that the mode of the predicted distribution outperform our regression baselines. The code for our implementation is available online.
☆ Intriguing Property of GAN for Remote Sensing Image Generation
Generative adversarial networks (GANs) have achieved remarkable progress in the natural image field. However, when applying GANs in the remote sensing (RS) image generation task, we discover an extraordinary phenomenon: the GAN model is more sensitive to the size of training data for RS image generation than for natural image generation. In other words, the generation quality of RS images will change significantly with the number of training categories or samples per category. In this paper, we first analyze this phenomenon from two kinds of toy experiments and conclude that the amount of feature information contained in the GAN model decreases with reduced training data. Based on this discovery, we propose two innovative adjustment schemes, namely Uniformity Regularization (UR) and Entropy Regularization (ER), to increase the information learned by the GAN model at the distributional and sample levels, respectively. We theoretically and empirically demonstrate the effectiveness and versatility of our methods. Extensive experiments on the NWPU-RESISC45 and PatternNet datasets show that our methods outperform the well-established models on RS image generation tasks.
☆ GPGait: Generalized Pose-based Gait Recognition
Recent works on pose-based gait recognition have demonstrated the potential of using such simple information to achieve results comparable to silhouette-based methods. However, the generalization ability of pose-based methods on different datasets is undesirably inferior to that of silhouette-based ones, which has received little attention but hinders the application of these methods in real-world scenarios. To improve the generalization ability of pose-based methods across datasets, we propose a Generalized Pose-based Gait recognition (GPGait) framework. First, a Human-Oriented Transformation (HOT) and a series of Human-Oriented Descriptors (HOD) are proposed to obtain a unified pose representation with discriminative multi-features. Then, given the slight variations in the unified representation after HOT and HOD, it becomes crucial for the network to extract local-global relationships between the keypoints. To this end, a Part-Aware Graph Convolutional Network (PAGCN) is proposed to enable efficient graph partition and local-global spatial feature extraction. Experiments on four public gait recognition datasets, CASIA-B, OUMVLP-Pose, Gait3D and GREW, show that our model demonstrates better and more stable cross-domain capabilities compared to existing skeleton-based methods, achieving comparable recognition results to silhouette-based ones. The code will be released.
comment: 10 pages, 4 figures
☆ Active Learning Based Domain Adaptation for Tissue Segmentation of Histopathological Images
Accurate segmentation of tissue in histopathological images can be very beneficial for defining regions of interest (ROI) for streamline of diagnostic and prognostic tasks. Still, adapting to different domains is essential for histopathology image analysis, as the visual characteristics of tissues can vary significantly across datasets. Yet, acquiring sufficient annotated data in the medical domain is cumbersome and time-consuming. The labeling effort can be significantly reduced by leveraging active learning, which enables the selective annotation of the most informative samples. Our proposed method allows for fine-tuning a pre-trained deep neural network using a small set of labeled data from the target domain, while also actively selecting the most informative samples to label next. We demonstrate that our approach performs with significantly fewer labeled samples compared to traditional supervised learning approaches for similar F1-scores, using barely a 59\% of the training set. We also investigate the distribution of class balance to establish annotation guidelines.
☆ Taming Contrast Maximization for Learning Sequential, Low-latency, Event-based Optical Flow
Event cameras have recently gained significant traction since they open up new avenues for low-latency and low-power solutions to complex computer vision problems. To unlock these solutions, it is necessary to develop algorithms that can leverage the unique nature of event data. However, the current state-of-the-art is still highly influenced by the frame-based literature, and usually fails to deliver on these promises. In this work, we take this into consideration and propose a novel self-supervised learning pipeline for the sequential estimation of event-based optical flow that allows for the scaling of the models to high inference frequencies. At its core, we have a continuously-running stateful neural model that is trained using a novel formulation of contrast maximization that makes it robust to nonlinearities and varying statistics in the input events. Results across multiple datasets confirm the effectiveness of our method, which establishes a new state of the art in terms of accuracy for approaches trained or optimized without ground truth.
comment: 15 pages, 12 figures, 7 tables
☆ Revisiting Rotation Averaging: Uncertainties and Robust Losses CVPR2023
In this paper, we revisit the rotation averaging problem applied in global Structure-from-Motion pipelines. We argue that the main problem of current methods is the minimized cost function that is only weakly connected with the input data via the estimated epipolar geometries.We propose to better model the underlying noise distributions by directly propagating the uncertainty from the point correspondences into the rotation averaging. Such uncertainties are obtained for free by considering the Jacobians of two-view refinements. Moreover, we explore integrating a variant of the MAGSAC loss into the rotation averaging problem, instead of using classical robust losses employed in current frameworks. The proposed method leads to results superior to baselines, in terms of accuracy, on large-scale public benchmarks. The code is public. https://github.com/zhangganlin/GlobalSfMpy
comment: submitted to CVPR2023
☆ Contrastive Model Adaptation for Cross-Condition Robustness in Semantic Segmentation
Standard unsupervised domain adaptation methods adapt models from a source to a target domain using labeled source data and unlabeled target data jointly. In model adaptation, on the other hand, access to the labeled source data is prohibited, i.e., only the source-trained model and unlabeled target data are available. We investigate normal-to-adverse condition model adaptation for semantic segmentation, whereby image-level correspondences are available in the target domain. The target set consists of unlabeled pairs of adverse- and normal-condition street images taken at GPS-matched locations. Our method -- CMA -- leverages such image pairs to learn condition-invariant features via contrastive learning. In particular, CMA encourages features in the embedding space to be grouped according to their condition-invariant semantic content and not according to the condition under which respective inputs are captured. To obtain accurate cross-domain semantic correspondences, we warp the normal image to the viewpoint of the adverse image and leverage warp-confidence scores to create robust, aggregated features. With this approach, we achieve state-of-the-art semantic segmentation performance for model adaptation on several normal-to-adverse adaptation benchmarks, such as ACDC and Dark Zurich. We also evaluate CMA on a newly procured adverse-condition generalization benchmark and report favorable results compared to standard unsupervised domain adaptation methods, despite the comparative handicap of CMA due to source data inaccessibility. Code is available at https://github.com/brdav/cma.
☆ Virtual Inverse Perspective Mapping for Simultaneous Pose and Motion Estimation
We propose an automatic method for pose and motion estimation against a ground surface for a ground-moving robot-mounted monocular camera. The framework adopts a semi-dense approach that benefits from both a feature-based method and an image-registration-based method by setting multiple patches in the image for displacement computation through a highly accurate image-registration technique. To improve accuracy, we introduce virtual inverse perspective mapping (IPM) in the refinement step to eliminate the perspective effect on image registration. The pose and motion are jointly and robustly estimated by a formulation of geometric bundle adjustment via virtual IPM. Unlike conventional visual odometry methods, the proposed method is free from cumulative error because it directly estimates pose and motion against the ground by taking advantage of a camera configuration mounted on a ground-moving robot where the camera's vertical motion is ignorable compared to its height within the frame interval and the nearby ground surface is approximately flat. We conducted experiments in which the relative mean error of the pitch and roll angles was approximately 1.0 degrees and the absolute mean error of the travel distance was 0.3 mm, even under camera shaking within a short period.
☆ Blind2Sound: Self-Supervised Image Denoising without Residual Noise
Self-supervised blind denoising for Poisson-Gaussian noise remains a challenging task. Pseudo-supervised pairs constructed from single noisy images re-corrupt the signal and degrade the performance. The visible blindspots solve the information loss in masked inputs. However, without explicitly noise sensing, mean square error as an objective function cannot adjust denoising intensities for dynamic noise levels, leading to noticeable residual noise. In this paper, we propose Blind2Sound, a simple yet effective approach to overcome residual noise in denoised images. The proposed adaptive re-visible loss senses noise levels and performs personalized denoising without noise residues while retaining the signal lossless. The theoretical analysis of intermediate medium gradients guarantees stable training, while the Cramer Gaussian loss acts as a regularization to facilitate the accurate perception of noise levels and improve the performance of the denoiser. Experiments on synthetic and real-world datasets show the superior performance of our method, especially for single-channel images.
☆ Classification in Histopathology: A unique deep embeddings extractor for multiple classification tasks
In biomedical imaging, deep learning-based methods are state-of-the-art for every modality (virtual slides, MRI, etc.) In histopathology, these methods can be used to detect certain biomarkers or classify lesions. However, such techniques require large amounts of data to train high-performing models which can be intrinsically difficult to acquire, especially when it comes to scarce biomarkers. To address this challenge, we use a single, pre-trained, deep embeddings extractor to convert images into deep features and train small, dedicated classification head on these embeddings for each classification task. This approach offers several benefits such as the ability to reuse a single pre-trained deep network for various tasks; reducing the amount of labeled data needed as classification heads have fewer parameters; and accelerating training time by up to 1000 times, which allows for much more tuning of the classification head. In this work, we perform an extensive comparison of various open-source backbones and assess their fit to the target histological image domain. This is achieved using a novel method based on a proxy classification task. We demonstrate that thanks to this selection method, an optimal feature extractor can be selected for different tasks on the target domain. We also introduce a feature space augmentation strategy which proves to substantially improve the final metrics computed for the different tasks considered. To demonstrate the benefit of such backbone selection and feature-space augmentation, our experiments are carried out on three separate classification tasks and show a clear improvement on each of them: microcalcifications (29.1% F1-score increase), lymph nodes metastasis (12.5% F1-score increase), mitosis (15.0% F1-score increase).
☆ RiDDLE: Reversible and Diversified De-identification with Latent Encryptor
This work presents RiDDLE, short for Reversible and Diversified De-identification with Latent Encryptor, to protect the identity information of people from being misused. Built upon a pre-learned StyleGAN2 generator, RiDDLE manages to encrypt and decrypt the facial identity within the latent space. The design of RiDDLE has three appealing properties. First, the encryption process is cipher-guided and hence allows diverse anonymization using different passwords. Second, the true identity can only be decrypted with the correct password, otherwise the system will produce another de-identified face to maintain the privacy. Third, both encryption and decryption share an efficient implementation, benefiting from a carefully tailored lightweight encryptor. Comparisons with existing alternatives confirm that our approach accomplishes the de-identification task with better quality, higher diversity, and stronger reversibility. We further demonstrate the effectiveness of RiDDLE in anonymizing videos. Code and models will be made publicly available.
☆ TAEC: Unsupervised Action Segmentation with Temporal-Aware Embedding and Clustering
Temporal action segmentation in untrimmed videos has gained increased attention recently. However, annotating action classes and frame-wise boundaries is extremely time consuming and cost intensive, especially on large-scale datasets. To address this issue, we propose an unsupervised approach for learning action classes from untrimmed video sequences. In particular, we propose a temporal embedding network that combines relative time prediction, feature reconstruction, and sequence-to-sequence learning, to preserve the spatial layout and sequential nature of the video features. A two-step clustering pipeline on these embedded feature representations then allows us to enforce temporal consistency within, as well as across videos. Based on the identified clusters, we decode the video into coherent temporal segments that correspond to semantically meaningful action classes. Our evaluation on three challenging datasets shows the impact of each component and, furthermore, demonstrates our state-of-the-art unsupervised action segmentation results.
comment: Computer Vision Winter Workshop 2023
☆ Reliability-Adaptive Consistency Regularization for Weakly-Supervised Point Cloud Segmentation
Weakly-supervised point cloud segmentation with extremely limited labels is highly desirable to alleviate the expensive costs of collecting densely annotated 3D points. This paper explores to apply the consistency regularization that is commonly used in weakly-supervised learning, for its point cloud counterpart with multiple data-specific augmentations, which has not been well studied. We observe that the straightforward way of applying consistency constraints to weakly-supervised point cloud segmentation has two major limitations: noisy pseudo labels due to the conventional confidence-based selection and insufficient consistency constraints due to discarding unreliable pseudo labels. Therefore, we propose a novel Reliability-Adaptive Consistency Network (RAC-Net) to use both prediction confidence and model uncertainty to measure the reliability of pseudo labels and apply consistency training on all unlabeled points while with different consistency constraints for different points based on the reliability of corresponding pseudo labels. Experimental results on the S3DIS and ScanNet-v2 benchmark datasets show that our model achieves superior performance in weakly-supervised point cloud segmentation. The code will be released.
☆ EVOLIN Benchmark: Evaluation of Line Detection and Association
Lines are interesting geometrical features commonly seen in indoor and urban environments. There is missing a complete benchmark where one can evaluate lines from a sequential stream of images in all its stages: Line detection, Line Association and Pose error. To do so, we present a complete and exhaustive benchmark for visual lines in a SLAM front-end, both for RGB and RGBD, by providing a plethora of complementary metrics. We have also labelled data from well-known SLAM datasets in order to have all in one poses and accurately annotated lines. In particular, we have evaluated 17 line detection algorithms, 5 line associations methods and the resultant pose error for aligning a pair of frames with several combinations of detector-association. We have packaged all methods and evaluations metrics and made them publicly available on web-page https://prime-slam.github.io/evolin/.
☆ Local Implicit Normalizing Flow for Arbitrary-Scale Image Super-Resolution CVPR 2023
Flow-based methods have demonstrated promising results in addressing the ill-posed nature of super-resolution (SR) by learning the distribution of high-resolution (HR) images with the normalizing flow. However, these methods can only perform a predefined fixed-scale SR, limiting their potential in real-world applications. Meanwhile, arbitrary-scale SR has gained more attention and achieved great progress. Nonetheless, previous arbitrary-scale SR methods ignore the ill-posed problem and train the model with per-pixel L1 loss, leading to blurry SR outputs. In this work, we propose "Local Implicit Normalizing Flow" (LINF) as a unified solution to the above problems. LINF models the distribution of texture details under different scaling factors with normalizing flow. Thus, LINF can generate photo-realistic HR images with rich texture details in arbitrary scale factors. We evaluate LINF with extensive experiments and show that LINF achieves the state-of-the-art perceptual quality compared with prior arbitrary-scale SR methods.
comment: Accepted to CVPR 2023
☆ 3D wind field profiles from hyperspectral sounders: revisiting optic-flow from a meteorological perspective
In this work, we present an efficient optic flow algorithm for the extraction of vertically resolved 3D atmospheric motion vector (AMV) fields from incomplete hyperspectral image data measures by infrared sounders. The model at the heart of the energy to be minimized is consistent with atmospheric dynamics, incorporating ingredients of thermodynamics, hydrostatic equilibrium and statistical turbulence. Modern optimization techniques are deployed to design a low-complexity solver for the energy minimization problem, which is non-convex, non-differentiable, high-dimensional and subject to physical constraints. In particular, taking advantage of the alternate direction of multipliers methods (ADMM), we show how to split the original high-dimensional problem into a recursion involving a set of standard and tractable optic-flow sub-problems. By comparing with the ground truth provided by the operational numerical simulation of the European Centre for Medium-Range Weather Forecasts (ECMWF), we show that the performance of the proposed method is superior to state-of-the-art optical flow algorithms in the context of real infrared atmospheric sounding interferometer (IASI) observations.
☆ Weakly Supervised Knowledge Transfer with Probabilistic Logical Reasoning for Object Detection ICLR 2023
Training object detection models usually requires instance-level annotations, such as the positions and labels of all objects present in each image. Such supervision is unfortunately not always available and, more often, only image-level information is provided, also known as weak supervision. Recent works have addressed this limitation by leveraging knowledge from a richly annotated domain. However, the scope of weak supervision supported by these approaches has been very restrictive, preventing them to use all available information. In this work, we propose ProbKT, a framework based on probabilistic logical reasoning that allows to train object detection models with arbitrary types of weak supervision. We empirically show on different datasets that using all available information is beneficial as our ProbKT leads to significant improvement on target domain and better generalization compared to existing baselines. We also showcase the ability of our approach to handle complex logic statements as supervision signal.
comment: Accepted to ICLR 2023
☆ Blind deblurring of hyperspectral document images
Most computer vision and machine learning-based approaches for historical document analysis are tailored to grayscale or RGB images and thus, mostly exploit their spatial information. Multispectral (MS) and hyperspectral (HS) images contain, next to the spatial information, much richer spectral information than RGB images (usually spreading beyond the visible spectral range) that can facilitate more effective feature extraction, more accurate classification and recognition, and thus, improved analysis. Although utilization of rich spectral information can improve historical document analysis tremendously, there are still some potential limitations of HS imagery such as camera-induced noise and blur that require a carefully designed preprocessing step. Here, we propose novel blind HS image deblurring methods tailored to document images. We exploit a low-rank property of HS images (i.e., by projecting an HS image to a lower dimensional subspace) and utilize a text tailor image prior to performing a PSF estimation and deblurring of subspace components. The preliminary results show that the proposed approach gives good results over all spectral bands, removing successfully image artefacts introduced by blur and noise and significantly increasing the number of bands that can be used in further analysis.
comment: This project has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement No. 101026453. This work is published in the Lecture Notes in Computer Science book series (LNCS, volume 13373) as part of the Image Analysis and Processing, ICIAP 2022 Workshops
☆ Hybrid Dual Mean-Teacher Network With Double-Uncertainty Guidance for Semi-Supervised Segmentation of MRI Scans
Semi-supervised learning has made significant progress in medical image segmentation. However, existing methods primarily utilize information acquired from a single dimensionality (2D/3D), resulting in sub-optimal performance on challenging data, such as magnetic resonance imaging (MRI) scans with multiple objects and highly anisotropic resolution. To address this issue, we present a Hybrid Dual Mean-Teacher (HD-Teacher) model with hybrid, semi-supervised, and multi-task learning to achieve highly effective semi-supervised segmentation. HD-Teacher employs a 2D and a 3D mean-teacher network to produce segmentation labels and signed distance fields from the hybrid information captured in both dimensionalities. This hybrid learning mechanism allows HD-Teacher to combine the `best of both worlds', utilizing features extracted from either 2D, 3D, or both dimensions to produce outputs as it sees fit. Outputs from 2D and 3D teacher models are also dynamically combined, based on their individual uncertainty scores, into a single hybrid prediction, where the hybrid uncertainty is estimated. We then propose a hybrid regularization module to encourage both student models to produce results close to the uncertainty-weighted hybrid prediction. The hybrid uncertainty suppresses unreliable knowledge in the hybrid prediction, leaving only useful information to improve network performance further. Extensive experiments of binary and multi-class segmentation conducted on three MRI datasets demonstrate the effectiveness of the proposed framework. Code is available at https://github.com/ThisGame42/Hybrid-Teacher.
☆ Cones: Concept Neurons in Diffusion Models for Customized Generation
Human brains respond to semantic features of presented stimuli with different neurons. It is then curious whether modern deep neural networks admit a similar behavior pattern. Specifically, this paper finds a small cluster of neurons in a diffusion model corresponding to a particular subject. We call those neurons the concept neurons. They can be identified by statistics of network gradients to a stimulation connected with the given subject. The concept neurons demonstrate magnetic properties in interpreting and manipulating generation results. Shutting them can directly yield the related subject contextualized in different scenes. Concatenating multiple clusters of concept neurons can vividly generate all related concepts in a single image. A few steps of further fine-tuning can enhance the multi-concept capability, which may be the first to manage to generate up to four different subjects in a single image. For large-scale applications, the concept neurons are environmentally friendly as we only need to store a sparse cluster of int index instead of dense float32 values of the parameters, which reduces storage consumption by 90\% compared with previous subject-driven generation methods. Extensive qualitative and quantitative studies on diverse scenarios show the superiority of our method in interpreting and manipulating diffusion models.
☆ Dominating Set Database Selection for Visual Place Recognition
This paper presents an approach for creating a visual place recognition (VPR) database for localization in indoor environments from RGBD scanning sequences. The proposed approach is formulated as a minimization problem in terms of dominating set algorithm for graph, constructed from spatial information, and referred as DominatingSet. Our algorithm shows better scene coverage in comparison to other methodologies that are used for database creation. Also, we demonstrate that using DominatingSet, a database size could be up to 250-1400 times smaller than the original scanning sequence while maintaining a recall rate of more than 80% on testing sequences. We evaluated our algorithm on 7-scenes and BundleFusion datasets and an additionally recorded sequence in a highly repetitive office setting. In addition, the database selection can produce weakly-supervised labels for fine-tuning neural place recognition algorithms to particular settings, improving even more their accuracy. The paper also presents a fully automated pipeline for VPR database creation from RGBD scanning sequences, as well as a set of metrics for VPR database evaluation. The code and released data are available on our web-page~ -- https://prime-slam.github.io/place-recognition-db/
☆ R-Tuning: Regularized Prompt Tuning in Open-Set Scenarios
In realistic open-set scenarios where labels of a part of testing data are totally unknown, current prompt methods on vision-language (VL) models always predict the unknown classes as the downstream training classes. The exhibited label bias causes difficulty in the open set recognition (OSR), by which an image should be correctly predicted as one of the known classes or the unknown one. To learn prompts in open-set scenarios, we propose the Regularized prompt Tuning (R-Tuning) to mitigate the label bias. It introduces open words from the WordNet to extend the range of words forming the prompt texts from only closed-set label words to more. Thus, prompts are tuned in a simulated open-set scenario. Besides, inspired by the observation that classifying directly on large datasets causes a much higher false positive rate than on small datasets, we propose the Combinatorial Tuning and Testing (CTT) strategy for improving performance. CTT decomposes R-Tuning on large datasets as multiple independent group-wise tuning on fewer classes, then makes comprehensive predictions by selecting the optimal sub-prompt. For fair comparisons, we construct new baselines for OSR based on VL models, especially for prompt methods. Our method achieves the best results on datasets with various scales. Extensive ablation studies validate the effectiveness of our method.
☆ SLCA: Slow Learner with Classifier Alignment for Continual Learning on a Pre-trained Model
The goal of continual learning is to improve the performance of recognition models in learning sequentially arrived data. Although most existing works are established on the premise of learning from scratch, growing efforts have been devoted to incorporating the benefits of pre-training. However, how to adaptively exploit the pre-trained knowledge for each incremental task while maintaining its generalizability remains an open question. In this work, we present an extensive analysis for continual learning on a pre-trained model (CLPM), and attribute the key challenge to a progressive overfitting problem. Observing that selectively reducing the learning rate can almost resolve this issue in the representation layer, we propose a simple but extremely effective approach named Slow Learner with Classifier Alignment (SLCA), which further improves the classification layer by modeling the class-wise distributions and aligning the classification layers in a post-hoc fashion. Across a variety of scenarios, our proposal provides substantial improvements for CLPM (e.g., up to 49.76%, 50.05%, 44.69% and 40.16% on Split CIFAR-100, Split ImageNet-R, Split CUB-200 and Split Cars-196, respectively), and thus outperforms state-of-the-art approaches by a large margin. Based on such a strong baseline, critical factors and promising directions are analyzed in-depth to facilitate subsequent research.
comment: Tech report. 11 pages, 8 figures
☆ Multi-level Memory-augmented Appearance-Motion Correspondence Framework for Video Anomaly Detection
Frame prediction based on AutoEncoder plays a significant role in unsupervised video anomaly detection. Ideally, the models trained on the normal data could generate larger prediction errors of anomalies. However, the correlation between appearance and motion information is underutilized, which makes the models lack an understanding of normal patterns. Moreover, the models do not work well due to the uncontrollable generalizability of deep AutoEncoder. To tackle these problems, we propose a multi-level memory-augmented appearance-motion correspondence framework. The latent correspondence between appearance and motion is explored via appearance-motion semantics alignment and semantics replacement training. Besides, we also introduce a Memory-Guided Suppression Module, which utilizes the difference from normal prototype features to suppress the reconstruction capacity caused by skip-connection, achieving the tradeoff between the good reconstruction of normal data and the poor reconstruction of abnormal data. Experimental results show that our framework outperforms the state-of-the-art methods, achieving AUCs of 99.6\%, 93.8\%, and 76.3\% on UCSD Ped2, CUHK Avenue, and ShanghaiTech datasets.
☆ Segmentation method for cerebral blood vessels from MRA using hysteresis
Segmentation of cerebral blood vessels from Magnetic Resonance Imaging (MRI) is an open problem that could be solved with deep learning (DL). However, annotated data for training is often scarce. Due to the absence of open-source tools, we aim to develop a classical segmentation method that generates vessel ground truth from Magnetic Resonance Angiography for DL training of segmentation across a variety of modalities. The method combines size-specific Hessian filters, hysteresis thresholding and connected component correction. The optimal choice of processing steps was evaluated with a blinded scoring by a clinician using 24 3D images. The results show that all method steps are necessary to produce the highest (14.2/15) vessel segmentation quality score. Omitting the connected component correction caused the largest quality loss. The method, which is available on GitHub, can be used to train DL models for vessel segmentation.
☆ Synthetic Pseudo Anomalies for Unsupervised Video Anomaly Detection: A Simple yet Efficient Framework based on Masked Autoencoder ICASSP2023
Due to the limited availability of anomalous samples for training, video anomaly detection is commonly viewed as a one-class classification problem. Many prevalent methods investigate the reconstruction difference produced by AutoEncoders (AEs) under the assumption that the AEs would reconstruct the normal data well while reconstructing anomalies poorly. However, even with only normal data training, the AEs often reconstruct anomalies well, which depletes their anomaly detection performance. To alleviate this issue, we propose a simple yet efficient framework for video anomaly detection. The pseudo anomaly samples are introduced, which are synthesized from only normal data by embedding random mask tokens without extra data processing. We also propose a normalcy consistency training strategy that encourages the AEs to better learn the regular knowledge from normal and corresponding pseudo anomaly data. This way, the AEs learn more distinct reconstruction boundaries between normal and abnormal data, resulting in superior anomaly discrimination capability. Experimental results demonstrate the effectiveness of the proposed method.
comment: Accepted to ICASSP2023
☆ Retinal Image Segmentation with Small Datasets
Many eye diseases like Diabetic Macular Edema (DME), Age-related Macular Degeneration (AMD), and Glaucoma manifest in the retina, can cause irreversible blindness or severely impair the central version. The Optical Coherence Tomography (OCT), a 3D scan of the retina with high qualitative information about the retinal morphology, can be used to diagnose and monitor changes in the retinal anatomy. Many Deep Learning (DL) methods have shared the success of developing an automated tool to monitor pathological changes in the retina. However, the success of these methods depend mainly on large datasets. To address the challenge from very small and limited datasets, we proposed a DL architecture termed CoNet (Coherent Network) for joint segmentation of layers and fluids in retinal OCT images on very small datasets (less than a hundred training samples). The proposed model was evaluated on the publicly available Duke DME dataset consisting of 110 B-Scans from 10 patients suffering from DME. Experimental results show that the proposed model outperformed both the human experts' annotation and the current state-of-the-art architectures by a clear margin with a mean Dice Score of 88% when trained on 55 images without any data augmentation.
comment: Submitted to Bioimaging 2023
☆ Updated version: A Video Anomaly Detection Framework based on Appearance-Motion Semantics Representation Consistency ICASSP2023
Video anomaly detection is an essential but challenging task. The prevalent methods mainly investigate the reconstruction difference between normal and abnormal patterns but ignore the semantics consistency between appearance and motion information of behavior patterns, making the results highly dependent on the local context of frame sequences and lacking the understanding of behavior semantics. To address this issue, we propose a framework of Appearance-Motion Semantics Representation Consistency that uses the gap of appearance and motion semantic representation consistency between normal and abnormal data. The two-stream structure is designed to encode the appearance and motion information representation of normal samples, and a novel consistency loss is proposed to enhance the consistency of feature semantics so that anomalies with low consistency can be identified. Moreover, the lower consistency features of anomalies can be used to deteriorate the quality of the predicted frame, which makes anomalies easier to spot. Experimental results demonstrate the effectiveness of the proposed method.
comment: Accepted to ICASSP2023. arXiv admin note: substantial text overlap with arXiv:2204.04151
☆ MaskDiff: Modeling Mask Distribution with Diffusion Probabilistic Model for Few-Shot Instance Segmentation
Few-shot instance segmentation extends the few-shot learning paradigm to the instance segmentation task, which tries to segment instance objects from a query image with a few annotated examples of novel categories. Conventional approaches have attempted to address the task via prototype learning, known as point estimation. However, this mechanism is susceptible to noise and suffers from bias due to a significant scarcity of data. To overcome the disadvantages of the point estimation mechanism, we propose a novel approach, dubbed MaskDiff, which models the underlying conditional distribution of a binary mask, which is conditioned on an object region and $K$-shot information. Inspired by augmentation approaches that perturb data with Gaussian noise for populating low data density regions, we model the mask distribution with a diffusion probabilistic model. In addition, we propose to utilize classifier-free guided mask sampling to integrate category information into the binary mask generation process. Without bells and whistles, our proposed method consistently outperforms state-of-the-art methods on both base and novel classes of the COCO dataset while simultaneously being more stable than existing methods.
☆ StyleDiff: Attribute Comparison Between Unlabeled Datasets in Latent Disentangled Space
One major challenge in machine learning applications is coping with mismatches between the datasets used in the development and those obtained in real-world applications. These mismatches may lead to inaccurate predictions and errors, resulting in poor product quality and unreliable systems. In this study, we propose StyleDiff to inform developers of the differences between the two datasets for the steady development of machine learning systems. Using disentangled image spaces obtained from recently proposed generative models, StyleDiff compares the two datasets by focusing on attributes in the images and provides an easy-to-understand analysis of the differences between the datasets. The proposed StyleDiff performs in $O (d N\log N)$, where $N$ is the size of the datasets and $d$ is the number of attributes, enabling the application to large datasets. We demonstrate that StyleDiff accurately detects differences between datasets and presents them in an understandable format using, for example, driving scenes datasets.
comment: 23 pages, 16 figures, under review
☆ Trajectory-Aware Body Interaction Transformer for Multi-Person Pose Forecasting CVPR2023
Multi-person pose forecasting remains a challenging problem, especially in modeling fine-grained human body interaction in complex crowd scenarios. Existing methods typically represent the whole pose sequence as a temporal series, yet overlook interactive influences among people based on skeletal body parts. In this paper, we propose a novel Trajectory-Aware Body Interaction Transformer (TBIFormer) for multi-person pose forecasting via effectively modeling body part interactions. Specifically, we construct a Temporal Body Partition Module that transforms all the pose sequences into a Multi-Person Body-Part sequence to retain spatial and temporal information based on body semantics. Then, we devise a Social Body Interaction Self-Attention (SBI-MSA) module, utilizing the transformed sequence to learn body part dynamics for inter- and intra-individual interactions. Furthermore, different from prior Euclidean distance-based spatial encodings, we present a novel and efficient Trajectory-Aware Relative Position Encoding for SBI-MSA to offer discriminative spatial information and additional interactive clues. On both short- and long-term horizons, we empirically evaluate our framework on CMU-Mocap, MuPoTS-3D as well as synthesized datasets (6 ~ 10 persons), and demonstrate that our method greatly outperforms the state-of-the-art methods. Code will be made publicly available upon acceptance.
comment: Accepted by CVPR2023, 10 pages, 8 figures. arXiv admin note: text overlap with arXiv:2208.09224
☆ Improving Video Retrieval by Adaptive Margin SIGIR 2021
Video retrieval is becoming increasingly important owing to the rapid emergence of videos on the Internet. The dominant paradigm for video retrieval learns video-text representations by pushing the distance between the similarity of positive pairs and that of negative pairs apart from a fixed margin. However, negative pairs used for training are sampled randomly, which indicates that the semantics between negative pairs may be related or even equivalent, while most methods still enforce dissimilar representations to decrease their similarity. This phenomenon leads to inaccurate supervision and poor performance in learning video-text representations. While most video retrieval methods overlook that phenomenon, we propose an adaptive margin changed with the distance between positive and negative pairs to solve the aforementioned issue. First, we design the calculation framework of the adaptive margin, including the method of distance measurement and the function between the distance and the margin. Then, we explore a novel implementation called "Cross-Modal Generalized Self-Distillation" (CMGSD), which can be built on the top of most video retrieval models with few modifications. Notably, CMGSD adds few computational overheads at train time and adds no computational overhead at test time. Experimental results on three widely used datasets demonstrate that the proposed method can yield significantly better performance than the corresponding backbone model, and it outperforms state-of-the-art methods by a large margin.
comment: Accepted by SIGIR 2021
☆ DDS3D: Dense Pseudo-Labels with Dynamic Threshold for Semi-Supervised 3D Object Detection ICRA
In this paper, we present a simple yet effective semi-supervised 3D object detector named DDS3D. Our main contributions have two-fold. On the one hand, different from previous works using Non-Maximal Suppression (NMS) or its variants for obtaining the sparse pseudo labels, we propose a dense pseudo-label generation strategy to get dense pseudo-labels, which can retain more potential supervision information for the student network. On the other hand, instead of traditional fixed thresholds, we propose a dynamic threshold manner to generate pseudo-labels, which can guarantee the quality and quantity of pseudo-labels during the whole training process. Benefiting from these two components, our DDS3D outperforms the state-of-the-art semi-supervised 3d object detection with mAP of 3.1% on the pedestrian and 2.1% on the cyclist under the same configuration of 1% samples. Extensive ablation studies on the KITTI dataset demonstrate the effectiveness of our DDS3D. The code and models will be made publicly available at https://github.com/hust-jy/DDS3D
comment: Accepted for publication in 2023 IEEE International Conference on Robotics and Automation (ICRA)
☆ Efficient Transformer-based 3D Object Detection with Dynamic Token Halting
Balancing efficiency and accuracy is a long-standing problem for deploying deep learning models. The trade-off is even more important for real-time safety-critical systems like autonomous vehicles. In this paper, we propose an effective approach for accelerating transformer-based 3D object detectors by dynamically halting tokens at different layers depending on their contribution to the detection task. Although halting a token is a non-differentiable operation, our method allows for differentiable end-to-end learning by leveraging an equivalent differentiable forward-pass. Furthermore, our framework allows halted tokens to be reused to inform the model's predictions through a straightforward token recycling mechanism. Our method significantly improves the Pareto frontier of efficiency versus accuracy when compared with the existing approaches. By halting tokens and increasing model capacity, we are able to improve the baseline model's performance without increasing the model's latency on the Waymo Open Dataset.
☆ Learning the Legibility of Visual Text Perturbations EACL 2023
Many adversarial attacks in NLP perturb inputs to produce visually similar strings ('ergo' $\rightarrow$ '$\epsilon$rgo') which are legible to humans but degrade model performance. Although preserving legibility is a necessary condition for text perturbation, little work has been done to systematically characterize it; instead, legibility is typically loosely enforced via intuitions around the nature and extent of perturbations. Particularly, it is unclear to what extent can inputs be perturbed while preserving legibility, or how to quantify the legibility of a perturbed string. In this work, we address this gap by learning models that predict the legibility of a perturbed string, and rank candidate perturbations based on their legibility. To do so, we collect and release \dataset, a human-annotated dataset comprising the legibility of visually perturbed text. Using this dataset, we build both text- and vision-based models which achieve up to $0.91$ F1 score in predicting whether an input is legible, and an accuracy of $0.86$ in predicting which of two given perturbations is more legible. Additionally, we discover that legible perturbations from the \dataset dataset are more effective at lowering the performance of NLP models than best-known attack strategies, suggesting that current models may be vulnerable to a broad range of perturbations beyond what is captured by existing visual attacks. Data, code, and models are available at https://github.com/dvsth/learning-legibility-2023.
comment: 9 pages, 11 figures. Long paper at EACL 2023
☆ GaitEditer: Attribute Editing for Gait Representation Learning
Gait pattern is a promising biometric for applications, as it can be captured from a distance without requiring individual cooperation. Nevertheless, existing gait datasets typically suffer from limited diversity, with indoor datasets requiring participants to walk along a fixed route in a restricted setting, and outdoor datasets containing only few walking sequences per subject. Prior generative methods have attempted to mitigate these limitations by building virtual gait datasets. They primarily focus on manipulating a single, specific gait attribute (e.g., viewpoint or carrying), and require the supervised data pairs for training, thus lacking the flexibility and diversity for practical usage. In contrast, our GaitEditer can act as an online module to edit a broad range of gait attributes, such as pants, viewpoint, and even age, in an unsupervised manner, which current gait generative methods struggle with. Additionally, GaitEidter also finely preserves both temporal continuity and identity characteristics in generated gait sequences. Experiments show that GaitEditer provides extensive knowledge for clothing-invariant and view-invariant gait representation learning under various challenging scenarios. The source code will be available.
☆ Learn More for Food Recognition via Progressive Self-Distillation
Food recognition has a wide range of applications, such as health-aware recommendation and self-service restaurants. Most previous methods of food recognition firstly locate informative regions in some weakly-supervised manners and then aggregate their features. However, location errors of informative regions limit the effectiveness of these methods to some extent. Instead of locating multiple regions, we propose a Progressive Self-Distillation (PSD) method, which progressively enhances the ability of network to mine more details for food recognition. The training of PSD simultaneously contains multiple self-distillations, in which a teacher network and a student network share the same embedding network. Since the student network receives a modified image from its teacher network by masking some informative regions, the teacher network outputs stronger semantic representations than the student network. Guided by such teacher network with stronger semantics, the student network is encouraged to mine more useful regions from the modified image by enhancing its own ability. The ability of the teacher network is also enhanced with the shared embedding network. By using progressive training, the teacher network incrementally improves its ability to mine more discriminative regions. In inference phase, only the teacher network is used without the help of the student network. Extensive experiments on three datasets demonstrate the effectiveness of our proposed method and state-of-the-art performance.
☆ Identification of Systematic Errors of Image Classifiers on Rare Subgroups
Despite excellent average-case performance of many image classifiers, their performance can substantially deteriorate on semantically coherent subgroups of the data that were under-represented in the training data. These systematic errors can impact both fairness for demographic minority groups as well as robustness and safety under domain shift. A major challenge is to identify such subgroups with subpar performance when the subgroups are not annotated and their occurrence is very rare. We leverage recent advances in text-to-image models and search in the space of textual descriptions of subgroups ("prompts") for subgroups where the target model has low performance on the prompt-conditioned synthesized data. To tackle the exponentially growing number of subgroups, we employ combinatorial testing. We denote this procedure as PromptAttack as it can be interpreted as an adversarial attack in a prompt space. We study subgroup coverage and identifiability with PromptAttack in a controlled setting and find that it identifies systematic errors with high accuracy. Thereupon, we apply PromptAttack to ImageNet classifiers and identify novel systematic errors on rare subgroups.
☆ MBPTrack: Improving 3D Point Cloud Tracking with Memory Networks and Box Priors
3D single object tracking has been a crucial problem for decades with numerous applications such as autonomous driving. Despite its wide-ranging use, this task remains challenging due to the significant appearance variation caused by occlusion and size differences among tracked targets. To address these issues, we present MBPTrack, which adopts a Memory mechanism to utilize past information and formulates localization in a coarse-to-fine scheme using Box Priors given in the first frame. Specifically, past frames with targetness masks serve as an external memory, and a transformer-based module propagates tracked target cues from the memory to the current frame. To precisely localize objects of all sizes, MBPTrack first predicts the target center via Hough voting. By leveraging box priors given in the first frame, we adaptively sample reference points around the target center that roughly cover the target of different sizes. Then, we obtain dense feature maps by aggregating point features into the reference points, where localization can be performed more effectively. Extensive experiments demonstrate that MBPTrack achieves state-of-the-art performance on KITTI, nuScenes and Waymo Open Dataset, while running at 50 FPS on a single RTX3090 GPU.
☆ Toward Unsupervised Realistic Visual Question Answering
The problem of realistic VQA (RVQA), where a model has to reject unanswerable questions (UQs) and answer answerable ones (AQs), is studied. We first point out 2 drawbacks in current RVQA research, where (1) datasets contain too many unchallenging UQs and (2) a large number of annotated UQs are required for training. To resolve the first drawback, we propose a new testing dataset, RGQA, which combines AQs from an existing VQA dataset with around 29K human-annotated UQs. These UQs consist of both fine-grained and coarse-grained image-question pairs generated with 2 approaches: CLIP-based and Perturbation-based. To address the second drawback, we introduce an unsupervised training approach. This combines pseudo UQs obtained by randomly pairing images and questions, with an RoI Mixup procedure to generate more fine-grained pseudo UQs, and model ensembling to regularize model confidence. Experiments show that using pseudo UQs significantly outperforms RVQA baselines. RoI Mixup and model ensembling further increase the gain. Finally, human evaluation reveals a performance gap between humans and models, showing that more RVQA research is needed.
comment: Yuwei Zhang and Chih-Hui Ho contributed equally to this work
☆ Distortion-Disentangled Contrastive Learning
Self-supervised learning is well known for its remarkable performance in representation learning and various downstream computer vision tasks. Recently, Positive-pair-Only Contrastive Learning (POCL) has achieved reliable performance without the need to construct positive-negative training sets. It reduces memory requirements by lessening the dependency on the batch size. The POCL method typically uses a single loss function to extract the distortion invariant representation (DIR) which describes the proximity of positive-pair representations affected by different distortions. This loss function implicitly enables the model to filter out or ignore the distortion variant representation (DVR) affected by different distortions. However, existing POCL methods do not explicitly enforce the disentanglement and exploitation of the actually valuable DVR. In addition, these POCL methods have been observed to be sensitive to augmentation strategies. To address these limitations, we propose a novel POCL framework named Distortion-Disentangled Contrastive Learning (DDCL) and a Distortion-Disentangled Loss (DDL). Our approach is the first to explicitly disentangle and exploit the DVR inside the model and feature stream to improve the overall representation utilization efficiency, robustness and representation ability. Experiments carried out demonstrate the superiority of our framework to Barlow Twins and Simsiam in terms of convergence, representation quality, and robustness on several benchmark datasets.
☆ Lifelong-MonoDepth: Lifelong Learning for Multi-Domain Monocular Metric Depth Estimation
In recent years, monocular depth estimation (MDE) has gained significant progress in a data-driven learning fashion. Previous methods can infer depth maps for specific domains based on the paradigm of single-domain or joint-domain training with mixed data. However, they suffer from low scalability to new domains. In reality, target domains often dynamically change or increase, raising the requirement of incremental multi-domain/task learning. In this paper, we seek to enable lifelong learning for MDE, which performs cross-domain depth learning sequentially, to achieve high plasticity on a new domain and maintain good stability on original domains. To overcome significant domain gaps and enable scale-aware depth prediction, we design a lightweight multi-head framework that consists of a domain-shared encoder for feature extraction and domain-specific predictors for metric depth estimation. Moreover, given an input image, we propose an efficient predictor selection approach that automatically identifies the corresponding predictor for depth inference. Through extensive numerical studies, we show that the proposed method can achieve good efficiency, stability, and plasticity, leading the benchmarks by 8% to 15%.
☆ Unifying Layout Generation with a Decoupled Diffusion Model CVPR2023
Layout generation aims to synthesize realistic graphic scenes consisting of elements with different attributes including category, size, position, and between-element relation. It is a crucial task for reducing the burden on heavy-duty graphic design works for formatted scenes, e.g., publications, documents, and user interfaces (UIs). Diverse application scenarios impose a big challenge in unifying various layout generation subtasks, including conditional and unconditional generation. In this paper, we propose a Layout Diffusion Generative Model (LDGM) to achieve such unification with a single decoupled diffusion model. LDGM views a layout of arbitrary missing or coarse element attributes as an intermediate diffusion status from a completed layout. Since different attributes have their individual semantics and characteristics, we propose to decouple the diffusion processes for them to improve the diversity of training samples and learn the reverse process jointly to exploit global-scope contexts for facilitating generation. As a result, our LDGM can generate layouts either from scratch or conditional on arbitrary available attributes. Extensive qualitative and quantitative experiments demonstrate our proposed LDGM outperforms existing layout generation models in both functionality and performance.
comment: Accepted by CVPR2023
☆ Diversity-Measurable Anomaly Detection CVPR 2023
Reconstruction-based anomaly detection models achieve their purpose by suppressing the generalization ability for anomaly. However, diverse normal patterns are consequently not well reconstructed as well. Although some efforts have been made to alleviate this problem by modeling sample diversity, they suffer from shortcut learning due to undesired transmission of abnormal information. In this paper, to better handle the tradeoff problem, we propose Diversity-Measurable Anomaly Detection (DMAD) framework to enhance reconstruction diversity while avoid the undesired generalization on anomalies. To this end, we design Pyramid Deformation Module (PDM), which models diverse normals and measures the severity of anomaly by estimating multi-scale deformation fields from reconstructed reference to original input. Integrated with an information compression module, PDM essentially decouples deformation from prototypical embedding and makes the final anomaly score more reliable. Experimental results on both surveillance videos and industrial images demonstrate the effectiveness of our method. In addition, DMAD works equally well in front of contaminated data and anomaly-like normal samples.
comment: Accepted by CVPR 2023
☆ Generative Model-Based Attack on Learnable Image Encryption for Privacy-Preserving Deep Learning
In this paper, we propose a novel generative model-based attack on learnable image encryption methods proposed for privacy-preserving deep learning. Various learnable encryption methods have been studied to protect the sensitive visual information of plain images, and some of them have been investigated to be robust enough against all existing attacks. However, previous attacks on image encryption focus only on traditional cryptanalytic attacks or reverse translation models, so these attacks cannot recover any visual information if a block-scrambling encryption step, which effectively destroys global information, is applied. Accordingly, in this paper, generative models are explored to evaluate whether such models can restore sensitive visual information from encrypted images for the first time. We first point out that encrypted images have some similarity with plain images in the embedding space. By taking advantage of leaked information from encrypted images, we propose a guided generative model as an attack on learnable image encryption to recover personally identifiable visual information. We implement the proposed attack in two ways by utilizing two state-of-the-art generative models: a StyleGAN-based model and latent diffusion-based one. Experiments were carried out on the CelebA-HQ and ImageNet datasets. Results show that images reconstructed by the proposed method have perceptual similarities to plain images.
comment: arXiv admin note: text overlap with arXiv:2209.07953
☆ CoralStyleCLIP: Co-optimized Region and Layer Selection for Image Editing CVPR 2023
Edit fidelity is a significant issue in open-world controllable generative image editing. Recently, CLIP-based approaches have traded off simplicity to alleviate these problems by introducing spatial attention in a handpicked layer of a StyleGAN. In this paper, we propose CoralStyleCLIP, which incorporates a multi-layer attention-guided blending strategy in the feature space of StyleGAN2 for obtaining high-fidelity edits. We propose multiple forms of our co-optimized region and layer selection strategy to demonstrate the variation of time complexity with the quality of edits over different architectural intricacies while preserving simplicity. We conduct extensive experimental analysis and benchmark our method against state-of-the-art CLIP-based methods. Our findings suggest that CoralStyleCLIP results in high-quality edits while preserving the ease of use.
comment: CVPR 2023
☆ SSL^2: Self-Supervised Learning meets Semi-Supervised Learning: Multiple Sclerosis Segmentation in 7T-MRI from large-scale 3T-MRI
Automated segmentation of multiple sclerosis (MS) lesions from MRI scans is important to quantify disease progression. In recent years, convolutional neural networks (CNNs) have shown top performance for this task when a large amount of labeled data is available. However, the accuracy of CNNs suffers when dealing with few and/or sparsely labeled datasets. A potential solution is to leverage the information available in large public datasets in conjunction with a target dataset which only has limited labeled data. In this paper, we propose a training framework, SSL2 (self-supervised-semi-supervised), for multi-modality MS lesion segmentation with limited supervision. We adopt self-supervised learning to leverage the knowledge from large public 3T datasets to tackle the limitations of a small 7T target dataset. To leverage the information from unlabeled 7T data, we also evaluate state-of-the-art semi-supervised methods for other limited annotation settings, such as small labeled training size and sparse annotations. We use the shifted-window (Swin) transformer1 as our backbone network. The effectiveness of self-supervised and semi-supervised training strategies is evaluated in our in-house 7T MRI dataset. The results indicate that each strategy improves lesion segmentation for both limited training data size and for sparse labeling scenarios. The combined overall framework further improves the performance substantially compared to either of its components alone. Our proposed framework thus provides a promising solution for future data/label-hungry 7T MS studies.
comment: Accepted at the International Society for Optics and Photonics - Medical Imaging (SPIE-MI) 2023
☆ DiffusionDepth: Diffusion Denoising Approach for Monocular Depth Estimation
Monocular depth estimation is a challenging task that predicts the pixel-wise depth from a single 2D image. Current methods typically model this problem as a regression or classification task. We propose DiffusionDepth, a new approach that reformulates monocular depth estimation as a denoising diffusion process. It learns an iterative denoising process to `denoise' random depth distribution into a depth map with the guidance of monocular visual conditions. The process is performed in the latent space encoded by a dedicated depth encoder and decoder. Instead of diffusing ground truth (GT) depth, the model learns to reverse the process of diffusing the refined depth of itself into random depth distribution. This self-diffusion formulation overcomes the difficulty of applying generative models to sparse GT depth scenarios. The proposed approach benefits this task by refining depth estimation step by step, which is superior for generating accurate and highly detailed depth maps. Experimental results on KITTI and NYU-Depth-V2 datasets suggest that a simple yet efficient diffusion approach could reach state-of-the-art performance in both indoor and outdoor scenarios with acceptable inference time.
☆ Smooth and Stepwise Self-Distillation for Object Detection
Distilling the structured information captured in feature maps has contributed to improved results for object detection tasks, but requires careful selection of baseline architectures and substantial pre-training. Self-distillation addresses these limitations and has recently achieved state-of-the-art performance for object detection despite making several simplifying architectural assumptions. Building on this work, we propose Smooth and Stepwise Self-Distillation (SSSD) for object detection. Our SSSD architecture forms an implicit teacher from object labels and a feature pyramid network backbone to distill label-annotated feature maps using Jensen-Shannon distance, which is smoother than distillation losses used in prior work. We additionally add a distillation coefficient that is adaptively configured based on the learning rate. We extensively benchmark SSSD against a baseline and two state-of-the-art object detector architectures on the COCO dataset by varying the coefficients and backbone and detector networks. We demonstrate that SSSD achieves higher average precision in most experimental settings, is robust to a wide range of coefficients, and benefits from our stepwise distillation procedure.
☆ Towards Robust Image-in-Audio Deep Steganography
The field of steganography has experienced a surge of interest due to the recent advancements in AI-powered techniques, particularly in the context of multimodal setups that enable the concealment of signals within signals of a different nature. The primary objectives of all steganographic methods are to achieve perceptual transparency, robustness, and large embedding capacity - which often present conflicting goals that classical methods have struggled to reconcile. This paper extends and enhances an existing image-in-audio deep steganography method by focusing on improving its robustness. The proposed enhancements include modifications to the loss function, utilization of the Short-Time Fourier Transform (STFT), introduction of redundancy in the encoding process for error correction, and buffering of additional information in the pixel subconvolution operation. The results demonstrate that our approach outperforms the existing method in terms of robustness and perceptual transparency.
comment: 8 pages, 5 figures, 2 tables
☆ Rethinking Visual Prompt Learning as Masked Visual Token Modeling
Prompt learning has achieved great success in efficiently exploiting large-scale pre-trained models in natural language processing (NLP). It reformulates the downstream tasks as the generative pre-training ones, thus narrowing down the gap between them and improving the performance stably. However, when transferring it to the vision area, current visual prompt learning methods are all designed on discriminative pre-trained models, and there is also a lack of careful design to unify the forms of pre-training and downstream tasks. To explore prompt learning on the generative pre-trained visual model as well as keeping the task consistency, we propose Visual Prompt learning as masked visual Token Modeling (VPTM) to transform the downstream visual classification into the pre-trained masked visual token prediction. In addition, we develop the prototypical verbalizer for mapping the predicted visual token with implicit semantics to explicit downstream labels. To our best knowledge, VPTM is the first visual prompt method on the generative pre-trained visual model, and the first to achieve consistency between pre-training and downstream visual classification by task reformulation. Experiments show that VPTM outperforms other visual prompt methods and achieves excellent efficiency. Moreover, the task consistency of VPTM contributes to the robustness against prompt location, prompt length and prototype dimension, and could be deployed uniformly.
☆ Optimization-Based Eye Tracking using Deflectometric Information
Eye tracking is an important tool with a wide range of applications in Virtual, Augmented, and Mixed Reality (VR/AR/MR) technologies. State-of-the-art eye tracking methods are either reflection-based and track reflections of sparse point light sources, or image-based and exploit 2D features of the acquired eye image. In this work, we attempt to significantly improve reflection-based methods by utilizing pixel-dense deflectometric surface measurements in combination with optimization-based inverse rendering algorithms. Utilizing the known geometry of our deflectometric setup, we develop a differentiable rendering pipeline based on PyTorch3D that simulates a virtual eye under screen illumination. Eventually, we exploit the image-screen-correspondence information from the captured measurements to find the eye's rotation, translation, and shape parameters with our renderer via gradient descent. In general, our method does not require a specific pattern and can work with ordinary video frames of the main VR/AR/MR screen itself. We demonstrate real-world experiments with evaluated mean relative gaze errors below 0.45 degrees at a precision better than 0.11 degrees. Moreover, we show an improvement of 6X over a representative reflection-based state-of-the-art method in simulation.
☆ Text-Visual Prompting for Efficient 2D Temporal Video Grounding CVPR 2023
In this paper, we study the problem of temporal video grounding (TVG), which aims to predict the starting/ending time points of moments described by a text sentence within a long untrimmed video. Benefiting from fine-grained 3D visual features, the TVG techniques have achieved remarkable progress in recent years. However, the high complexity of 3D convolutional neural networks (CNNs) makes extracting dense 3D visual features time-consuming, which calls for intensive memory and computing resources. Towards efficient TVG, we propose a novel text-visual prompting (TVP) framework, which incorporates optimized perturbation patterns (that we call 'prompts') into both visual inputs and textual features of a TVG model. In sharp contrast to 3D CNNs, we show that TVP allows us to effectively co-train vision encoder and language encoder in a 2D TVG model and improves the performance of crossmodal feature fusion using only low-complexity sparse 2D visual features. The proposed prompts also compensate for the lack of spatiotemporal information in 2D CNNs for visual feature extraction. Further, we propose a TemporalDistance IoU (TDIoU) loss for efficient learning of TVG. Last but not least, extensive experiments on two benchmark datasets, Charades-STA and ActivityNet Captions datasets, empirically show that the proposed TVP significantly boosts the performance of 2D TVG (e.g., 9.79% improvement in Charades-STA and 30.77% improvement in ActivityNet Captions) and achieves 5x inference acceleration over TVG of using 3D visual features. Code and model will be released.
comment: Accepted into the CVPR 2023
☆ Deformer: Dynamic Fusion Transformer for Robust Hand Pose Estimation
Accurately estimating 3D hand pose is crucial for understanding how humans interact with the world. Despite remarkable progress, existing methods often struggle to generate plausible hand poses when the hand is heavily occluded or blurred. In videos, the movements of the hand allow us to observe various parts of the hand that may be occluded or blurred in a single frame. To adaptively leverage the visual clue before and after the occlusion or blurring for robust hand pose estimation, we propose the Deformer: a framework that implicitly reasons about the relationship between hand parts within the same image (spatial dimension) and different timesteps (temporal dimension). We show that a naive application of the transformer self-attention mechanism is not sufficient because motion blur or occlusions in certain frames can lead to heavily distorted hand features and generate imprecise keys and queries. To address this challenge, we incorporate a Dynamic Fusion Module into Deformer, which predicts the deformation of the hand and warps the hand mesh predictions from nearby frames to explicitly support the current frame estimation. Furthermore, we have observed that errors are unevenly distributed across different hand parts, with vertices around fingertips having disproportionately higher errors than those around the palm. We mitigate this issue by introducing a new loss function called maxMSE that automatically adjusts the weight of every vertex to focus the model on critical hand parts. Extensive experiments show that our method significantly outperforms state-of-the-art methods by 10%, and is more robust to occlusions (over 14%).
☆ ARS-DETR: Aspect Ratio Sensitive Oriented Object Detection with Transformer
Existing oriented object detection methods commonly use metric AP$_{50}$ to measure the performance of the model. We argue that AP$_{50}$ is inherently unsuitable for oriented object detection due to its large tolerance in angle deviation. Therefore, we advocate using high-precision metric, e.g. AP$_{75}$, to measure the performance of models. In this paper, we propose an Aspect Ratio Sensitive Oriented Object Detector with Transformer, termed ARS-DETR, which exhibits a competitive performance in high-precision oriented object detection. Specifically, a new angle classification method, calling Aspect Ratio aware Circle Smooth Label (AR-CSL), is proposed to smooth the angle label in a more reasonable way and discard the hyperparameter that introduced by previous work (e.g. CSL). Then, a rotated deformable attention module is designed to rotate the sampling points with the corresponding angles and eliminate the misalignment between region features and sampling points. Moreover, a dynamic weight coefficient according to the aspect ratio is adopted to calculate the angle loss. Comprehensive experiments on several challenging datasets show that our method achieves competitive performance on the high-precision oriented object detection task.
comment: 10 pages, 8 figures, 8 tables, the source code is available at https://github.com/httle/ARS-DETR
☆ Decision-BADGE: Decision-based Adversarial Batch Attack with Directional Gradient Estimation
The vulnerability of deep neural networks to adversarial examples has led to the rise in the use of adversarial attacks. While various decision-based and universal attack methods have been proposed, none have attempted to create a decision-based universal adversarial attack. This research proposes Decision-BADGE, which uses random gradient-free optimization and batch attack to generate universal adversarial perturbations for decision-based attacks. Multiple adversarial examples are combined to optimize a single universal perturbation, and the accuracy metric is reformulated into a continuous Hamming distance form. The effectiveness of accuracy metric as a loss function is demonstrated and mathematically proven. The combination of Decision-BADGE and the accuracy loss function performs better than both score-based image-dependent attack and white-box universal attack methods in terms of attack time efficiency. The research also shows that Decision-BADGE can successfully deceive unseen victims and accurately target specific classes.
comment: 10 pages (8 pages except for references), 6 figures, 6 tables
☆ Curvature-Sensitive Predictive Coding with Approximate Laplace Monte Carlo
Predictive coding (PC) accounts of perception now form one of the dominant computational theories of the brain, where they prescribe a general algorithm for inference and learning over hierarchical latent probabilistic models. Despite this, they have enjoyed little export to the broader field of machine learning, where comparative generative modelling techniques have flourished. In part, this has been due to the poor performance of models trained with PC when evaluated by both sample quality and marginal likelihood. By adopting the perspective of PC as a variational Bayes algorithm under the Laplace approximation, we identify the source of these deficits to lie in the exclusion of an associated Hessian term in the PC objective function, which would otherwise regularise the sharpness of the probability landscape and prevent over-certainty in the approximate posterior. To remedy this, we make three primary contributions: we begin by suggesting a simple Monte Carlo estimated evidence lower bound which relies on sampling from the Hessian-parameterised variational posterior. We then derive a novel block diagonal approximation to the full Hessian matrix that has lower memory requirements and favourable mathematical properties. Lastly, we present an algorithm that combines our method with standard PC to reduce memory complexity further. We evaluate models trained with our approach against the standard PC framework on image benchmark datasets. Our approach produces higher log-likelihoods and qualitatively better samples that more closely capture the diversity of the data-generating distribution.
☆ LMR: A Large-Scale Multi-Reference Dataset for Reference-based Super-Resolution
It is widely agreed that reference-based super-resolution (RefSR) achieves superior results by referring to similar high quality images, compared to single image super-resolution (SISR). Intuitively, the more references, the better performance. However, previous RefSR methods have all focused on single-reference image training, while multiple reference images are often available in testing or practical applications. The root cause of such training-testing mismatch is the absence of publicly available multi-reference SR training datasets, which greatly hinders research efforts on multi-reference super-resolution. To this end, we construct a large-scale, multi-reference super-resolution dataset, named LMR. It contains 112,142 groups of 300x300 training images, which is 10x of the existing largest RefSR dataset. The image size is also much larger. More importantly, each group is equipped with 5 reference images with different similarity levels. Furthermore, we propose a new baseline method for multi-reference super-resolution: MRefSR, including a Multi-Reference Attention Module (MAM) for feature fusion of an arbitrary number of reference images, and a Spatial Aware Filtering Module (SAFM) for the fused feature selection. The proposed MRefSR achieves significant improvements over state-of-the-art approaches on both quantitative and qualitative evaluations. Our code and data would be made available soon.
comment: 6 figures, 10 pages
☆ CFR-ICL: Cascade-Forward Refinement with Iterative Click Loss for Interactive Image Segmentation
The click-based interactive segmentation aims to extract the object of interest from an image with the guidance of user clicks. Recent work has achieved great overall performance by employing the segmentation from the previous output. However, in most state-of-the-art approaches, 1) the inference stage involves inflexible heuristic rules and a separate refinement model; and 2) the training cannot balance the number of user clicks and model performance. To address the challenges, we propose a click-based and mask-guided interactive image segmentation framework containing three novel components: Cascade-Forward Refinement (CFR), Iterative Click Loss (ICL), and SUEM image augmentation. The proposed ICL allows model training to improve segmentation and reduce user interactions simultaneously. The CFR offers a unified inference framework to generate segmentation results in a coarse-to-fine manner. The proposed SUEM augmentation is a comprehensive way to create large and diverse training sets for interactive image segmentation. Extensive experiments demonstrate the state-of-the-art performance of the proposed approach on five public datasets. Remarkably, our model achieves an average of 2.9 and 7.5 clicks of NoC@95 on the Berkeley and DAVIS sets, respectively, improving by 33.2% and 15.5% over the previous state-of-the-art results. The code and trained model are available at https://github.com/TitorX/CFR-ICL-Interactive-Segmentation.
☆ KGNv2: Separating Scale and Pose Prediction for Keypoint-based 6-DoF Grasp Pose Synthesis on RGB-D input IROS2023
We propose a new 6-DoF grasp pose synthesis approach from 2D/2.5D input based on keypoints. Keypoint-based grasp detector from image input has demonstrated promising results in the previous study, where the additional visual information provided by color images compensates for the noisy depth perception. However, it relies heavily on accurately predicting the location of keypoints in the image space. In this paper, we devise a new grasp generation network that reduces the dependency on precise keypoint estimation. Given an RGB-D input, our network estimates both the grasp pose from keypoint detection as well as scale towards the camera. We further re-design the keypoint output space in order to mitigate the negative impact of keypoint prediction noise to Perspective-n-Point (PnP) algorithm. Experiments show that the proposed method outperforms the baseline by a large margin, validating the efficacy of our approach. Finally, despite trained on simple synthetic objects, our method demonstrate sim-to-real capacity by showing competitive results in real-world robot experiments.
comment: Submitted to IROS2023
☆ An Evaluation of Non-Contrastive Self-Supervised Learning for Federated Medical Image Analysis
Privacy and annotation bottlenecks are two major issues that profoundly affect the practicality of machine learning-based medical image analysis. Although significant progress has been made in these areas, these issues are not yet fully resolved. In this paper, we seek to tackle these concerns head-on and systematically explore the applicability of non-contrastive self-supervised learning (SSL) algorithms under federated learning (FL) simulations for medical image analysis. We conduct thorough experimentation of recently proposed state-of-the-art non-contrastive frameworks under standard FL setups. With the SoTA Contrastive Learning algorithm, SimCLR as our comparative baseline, we benchmark the performances of our 4 chosen non-contrastive algorithms under non-i.i.d. data conditions and with a varying number of clients. We present a holistic evaluation of these techniques on 6 standardized medical imaging datasets. We further analyse different trends inferred from the findings of our research, with the aim to find directions for further research based on ours. To the best of our knowledge, ours is the first to perform such a thorough analysis of federated self-supervised learning for medical imaging. All of our source code will be made public upon acceptance of the paper.
☆ EfficientTempNet: Temporal Super-Resolution of Radar Rainfall ICLR 2023
Rainfall data collected by various remote sensing instruments such as radars or satellites has different space-time resolutions. This study aims to improve the temporal resolution of radar rainfall products to help with more accurate climate change modeling and studies. In this direction, we introduce a solution based on EfficientNetV2, namely EfficientTempNet, to increase the temporal resolution of radar-based rainfall products from 10 minutes to 5 minutes. We tested EfficientRainNet over a dataset for the state of Iowa, US, and compared its performance to three different baselines to show that EfficientTempNet presents a viable option for better climate change monitoring.
comment: Published as a workshop paper at Tackling Climate Change with Machine Learning, ICLR 2023
☆ Weakly-Supervised HOI Detection from Interaction Labels Only and Language/Vision-Language Priors
Human-object interaction (HOI) detection aims to extract interacting human-object pairs and their interaction categories from a given natural image. Even though the labeling effort required for building HOI detection datasets is inherently more extensive than for many other computer vision tasks, weakly-supervised directions in this area have not been sufficiently explored due to the difficulty of learning human-object interactions with weak supervision, rooted in the combinatorial nature of interactions over the object and predicate space. In this paper, we tackle HOI detection with the weakest supervision setting in the literature, using only image-level interaction labels, with the help of a pretrained vision-language model (VLM) and a large language model (LLM). We first propose an approach to prune non-interacting human and object proposals to increase the quality of positive pairs within the bag, exploiting the grounding capability of the vision-language model. Second, we use a large language model to query which interactions are possible between a human and a given object category, in order to force the model not to put emphasis on unlikely interactions. Lastly, we use an auxiliary weakly-supervised preposition prediction task to make our model explicitly reason about space. Extensive experiments and ablations show that all of our contributions increase HOI detection performance.
comment: 8 pages, 3 figures and 5 tables
♻ ☆ Ancilia: Scalable Intelligent Video Surveillance for the Artificial Intelligence of Things
With the advancement of vision-based artificial intelligence, the proliferation of the Internet of Things connected cameras, and the increasing societal need for rapid and equitable security, the demand for accurate real-time intelligent surveillance has never been higher. This article presents Ancilia, an end-to-end scalable, intelligent video surveillance system for the Artificial Intelligence of Things. Ancilia brings state-of-the-art artificial intelligence to real-world surveillance applications while respecting ethical concerns and performing high-level cognitive tasks in real-time. Ancilia aims to revolutionize the surveillance landscape, to bring more effective, intelligent, and equitable security to the field, resulting in safer and more secure communities without requiring people to compromise their right to privacy.
♻ ☆ Multimodal Prompting with Missing Modalities for Visual Recognition CVPR 2023
In this paper, we tackle two challenges in multimodal learning for visual recognition: 1) when missing-modality occurs either during training or testing in real-world situations; and 2) when the computation resources are not available to finetune on heavy transformer models. To this end, we propose to utilize prompt learning and mitigate the above two challenges together. Specifically, our modality-missing-aware prompts can be plugged into multimodal transformers to handle general missing-modality cases, while only requiring less than 1% learnable parameters compared to training the entire model. We further explore the effect of different prompt configurations and analyze the robustness to missing modality. Extensive experiments are conducted to show the effectiveness of our prompt learning framework that improves the performance under various missing-modality cases, while alleviating the requirement of heavy model re-training. Code is available.
comment: Accepted by CVPR 2023. Codes are available at https://github.com/YiLunLee/Missing_aware_prompts
♻ ☆ CHAD: Charlotte Anomaly Dataset
In recent years, we have seen a significant interest in data-driven deep learning approaches for video anomaly detection, where an algorithm must determine if specific frames of a video contain abnormal behaviors. However, video anomaly detection is particularly context-specific, and the availability of representative datasets heavily limits real-world accuracy. Additionally, the metrics currently reported by most state-of-the-art methods often do not reflect how well the model will perform in real-world scenarios. In this article, we present the Charlotte Anomaly Dataset (CHAD). CHAD is a high-resolution, multi-camera anomaly dataset in a commercial parking lot setting. In addition to frame-level anomaly labels, CHAD is the first anomaly dataset to include bounding box, identity, and pose annotations for each actor. This is especially beneficial for skeleton-based anomaly detection, which is useful for its lower computational demand in real-world settings. CHAD is also the first anomaly dataset to contain multiple views of the same scene. With four camera views and over 1.15 million frames, CHAD is the largest fully annotated anomaly detection dataset including person annotations, collected from continuous video streams from stationary cameras for smart video surveillance applications. To demonstrate the efficacy of CHAD for training and evaluation, we benchmark two state-of-the-art skeleton-based anomaly detection algorithms on CHAD and provide comprehensive analysis, including both quantitative results and qualitative examination. The dataset is available at https://github.com/TeCSAR-UNCC/CHAD.
♻ ☆ City3D: Large-Scale Building Reconstruction from Airborne LiDAR Point Clouds
We present a fully automatic approach for reconstructing compact 3D building models from large-scale airborne point clouds. A major challenge of urban reconstruction from airborne LiDAR point clouds lies in that the vertical walls are typically missing. Based on the observation that urban buildings typically consist of planar roofs connected with vertical walls to the ground, we propose an approach to infer the vertical walls directly from the data. With the planar segments of both roofs and walls, we hypothesize the faces of the building surface, and the final model is obtained by using an extended hypothesis-and-selection-based polygonal surface reconstruction framework. Specifically, we introduce a new energy term to encourage roof preferences and two additional hard constraints into the optimization step to ensure correct topology and enhance detail recovery. Experiments on various large-scale airborne LiDAR point clouds have demonstrated that the method is superior to the state-of-the-art methods in terms of reconstruction accuracy and robustness. In addition, we have generated a new dataset with our method consisting of the point clouds and 3D models of 20k real-world buildings. We believe this dataset can stimulate research in urban reconstruction from airborne LiDAR point clouds and the use of 3D city models in urban applications.
♻ ☆ ELODIN: Naming Concepts in Embedding Spaces
Despite recent advancements, the field of text-to-image synthesis still suffers from lack of fine-grained control. Using only text, it remains challenging to deal with issues such as concept coherence and concept contamination. We propose a method to enhance control by generating specific concepts that can be reused throughout multiple images, effectively expanding natural language with new words that can be combined much like a painter's palette. Unlike previous contributions, our method does not copy visuals from input data and can generate concepts through text alone. We perform a set of comparisons that finds our method to be a significant improvement over text-only prompts.
comment: Added quantitative data, fixed formatting issues
♻ ☆ Using Multiple Instance Learning to Build Multimodal Representations
Image-text multimodal representation learning aligns data across modalities and enables important medical applications, e.g., image classification, visual grounding, and cross-modal retrieval. In this work, we establish a connection between multimodal representation learning and multiple instance learning. Based on this connection, we propose a generic framework for constructing permutation-invariant score functions with many existing multimodal representation learning approaches as special cases. Furthermore, we use the framework to derive a novel contrastive learning approach and demonstrate that our method achieves state-of-the-art results in several downstream tasks.
♻ ☆ Three New Validators and a Large-Scale Benchmark Ranking for Unsupervised Domain Adaptation
Changes to hyperparameters can have a dramatic effect on model accuracy. Thus, the tuning of hyperparameters plays an important role in optimizing machine-learning models. An integral part of the hyperparameter-tuning process is the evaluation of model checkpoints, which is done through the use of "validators". In a supervised setting, these validators evaluate checkpoints by computing accuracy on a validation set that has labels. In contrast, in an unsupervised setting, the validation set has no such labels. Without any labels, it is impossible to compute accuracy, so validators must estimate accuracy instead. But what is the best approach to estimating accuracy? In this paper, we consider this question in the context of unsupervised domain adaptation (UDA). Specifically, we propose three new validators, and we compare and rank them against five other existing validators, on a large dataset of 1,000,000 checkpoints. Extensive experimental results show that two of our proposed validators achieve state-of-the-art performance in various settings. Finally, we find that in many cases, the state-of-the-art is obtained by a simple baseline method. To the best of our knowledge, this is the largest empirical study of UDA validators to date. Code is available at https://www.github.com/KevinMusgrave/powerful-benchmarker.
comment: This paper was previously titled Benchmarking Validation Methods for Unsupervised Domain Adaptation. This version contains new experiments, analysis, and figures
♻ ☆ Privacy Preserving Image Registration
Image registration is a key task in medical imaging applications, allowing to represent medical images in a common spatial reference frame. Current approaches to image registration are generally based on the assumption that the content of the images is usually accessible in clear form, from which the spatial transformation is subsequently estimated. This common assumption may not be met in practical applications, since the sensitive nature of medical images may ultimately require their analysis under privacy constraints, preventing to openly share the image content.In this work, we formulate the problem of image registration under a privacy preserving regime, where images are assumed to be confidential and cannot be disclosed in clear. We derive our privacy preserving image registration framework by extending classical registration paradigms to account for advanced cryptographic tools, such as secure multi-party computation and homomorphic encryption, that enable the execution of operations without leaking the underlying data. To overcome the problem of performance and scalability of cryptographic tools in high dimensions, we propose several techniques to optimize the image registration operations by using gradient approximations, and by revisiting the use of homomorphic encryption trough packing, to allow the efficient encryption and multiplication of large matrices. We demonstrate our privacy preserving framework in linear and non-linear registration problems, evaluating its accuracy and scalability with respect to standard, non-private counterparts. Our results show that privacy preserving image registration is feasible and can be adopted in sensitive medical imaging applications.
♻ ☆ Bootstrap The Original Latent: Learning a Private Model from a Black-box Model
In this paper, considering the balance of data/model privacy of model owners and user needs, we propose a new setting called Back-Propagated Black-Box Adaptation (BPBA) for users to better train their private models via the guidance of the back-propagated results of a Black-box foundation/source model. Our setting can ease the usage of foundation/source models as well as prevent the leakage and misuse of foundation/source models. Moreover, we also propose a new training strategy called Bootstrap The Original Latent (BTOL) to fully utilize the foundation/source models. Our strategy consists of a domain adapter and a freeze-and-thaw strategy. We apply our BTOL under BPBA and Black-box UDA settings on three different datasets. Experiments show that our strategy is efficient and robust in various settings without manual augmentations.
♻ ☆ Open-Source Tools for Behavioral Video Analysis: Setup, Methods, and Development
Recently developed methods for video analysis, especially models for pose estimation and behavior classification, are transforming behavioral quantification to be more precise, scalable, and reproducible in fields such as neuroscience and ethology. These tools overcome long-standing limitations of manual scoring of video frames and traditional "center of mass" tracking algorithms to enable video analysis at scale. The expansion of open-source tools for video acquisition and analysis has led to new experimental approaches to understand behavior. Here, we review currently available open-source tools for video analysis and discuss how to set up these methods for labs new to video recording. We also discuss best practices for developing and using video analysis methods, including community-wide standards and critical needs for the open sharing of datasets and code, more widespread comparisons of video analysis methods, and better documentation for these methods especially for new users. We encourage broader adoption and continued development of these tools, which have tremendous potential for accelerating scientific progress in understanding the brain and behavior.
comment: 26 pages, 2 figures, 3 tables; this is a commentary on video methods for analyzing behavior in animals that emerged from a working group organized by the OpenBehavior project (openbehavior.com)
♻ ☆ LidarCLIP or: How I Learned to Talk to Point Clouds
Research connecting text and images has recently seen several breakthroughs, with models like CLIP, DALL-E 2, and Stable Diffusion. However, the connection between text and other visual modalities, such as lidar data, has received less attention, prohibited by the lack of text-lidar datasets. In this work, we propose LidarCLIP, a mapping from automotive point clouds to a pre-existing CLIP embedding space. Using image-lidar pairs, we supervise a point cloud encoder with the image CLIP embeddings, effectively relating text and lidar data with the image domain as an intermediary. We show the effectiveness of LidarCLIP by demonstrating that lidar-based retrieval is generally on par with image-based retrieval, but with complementary strengths and weaknesses. By combining image and lidar features, we improve upon both single-modality methods and enable a targeted search for challenging detection scenarios under adverse sensor conditions. We also explore zero-shot classification and show that LidarCLIP outperforms existing attempts to use CLIP for point clouds by a large margin. Finally, we leverage our compatibility with CLIP to explore a range of applications, such as point cloud captioning and lidar-to-image generation, without any additional training. Code and pre-trained models are available at https://github.com/atonderski/lidarclip.
♻ ☆ DREAM: Efficient Dataset Distillation by Representative Matching
Dataset distillation aims to synthesize small datasets with little information loss from original large-scale ones for reducing storage and training costs. Recent state-of-the-art methods mainly constrain the sample synthesis process by matching synthetic images and the original ones regarding gradients, embedding distributions, or training trajectories. Although there are various matching objectives, currently the strategy for selecting original images is limited to naive random sampling. We argue that random sampling overlooks the evenness of the selected sample distribution, which may result in noisy or biased matching targets. Besides, the sample diversity is also not constrained by random sampling. These factors together lead to optimization instability in the distilling process and degrade the training efficiency. Accordingly, we propose a novel matching strategy named as \textbf{D}ataset distillation by \textbf{RE}present\textbf{A}tive \textbf{M}atching (DREAM), where only representative original images are selected for matching. DREAM is able to be easily plugged into popular dataset distillation frameworks and reduce the distilling iterations by more than 8 times without performance drop. Given sufficient training time, DREAM further provides significant improvements and achieves state-of-the-art performances.
comment: Efficient matching for dataset distillation
♻ ☆ Masked Autoencoder for Self-Supervised Pre-training on Lidar Point Clouds
Masked autoencoding has become a successful pretraining paradigm for Transformer models for text, images, and, recently, point clouds. Raw automotive datasets are suitable candidates for self-supervised pre-training as they generally are cheap to collect compared to annotations for tasks like 3D object detection (OD). However, the development of masked autoencoders for point clouds has focused solely on synthetic and indoor data. Consequently, existing methods have tailored their representations and models toward small and dense point clouds with homogeneous point densities. In this work, we study masked autoencoding for point clouds in an automotive setting, which are sparse and for which the point density can vary drastically among objects in the same scene. To this end, we propose Voxel-MAE, a simple masked autoencoding pre-training scheme designed for voxel representations. We pre-train the backbone of a Transformer-based 3D object detector to reconstruct masked voxels and to distinguish between empty and non-empty voxels. Our method improves the 3D OD performance by 1.75 mAP points and 1.05 NDS on the challenging nuScenes dataset. Further, we show that by pre-training with Voxel-MAE, we require only 40% of the annotated data to outperform a randomly initialized equivalent. Code available at https://github.com/georghess/voxel-mae
♻ ☆ RGB-Event Fusion for Moving Object Detection in Autonomous Driving ICRA'23
Moving Object Detection (MOD) is a critical vision task for successfully achieving safe autonomous driving. Despite plausible results of deep learning methods, most existing approaches are only frame-based and may fail to reach reasonable performance when dealing with dynamic traffic participants. Recent advances in sensor technologies, especially the Event camera, can naturally complement the conventional camera approach to better model moving objects. However, event-based works often adopt a pre-defined time window for event representation, and simply integrate it to estimate image intensities from events, neglecting much of the rich temporal information from the available asynchronous events. Therefore, from a new perspective, we propose RENet, a novel RGB-Event fusion Network, that jointly exploits the two complementary modalities to achieve more robust MOD under challenging scenarios for autonomous driving. Specifically, we first design a temporal multi-scale aggregation module to fully leverage event frames from both the RGB exposure time and larger intervals. Then we introduce a bi-directional fusion module to attentively calibrate and fuse multi-modal features. To evaluate the performance of our network, we carefully select and annotate a sub-MOD dataset from the commonly used DSEC dataset. Extensive experiments demonstrate that our proposed method performs significantly better than the state-of-the-art RGB-Event fusion alternatives. The source code and dataset are publicly available at: https://github.com/ZZY-Zhou/RENet.
comment: ICRA'23
♻ ☆ Energy-Latency Attacks via Sponge Poisoning
Sponge examples are test-time inputs carefully optimized to increase energy consumption and latency of neural networks when deployed on hardware accelerators. In this work, we are the first to demonstrate that sponge examples can also be injected at training time, via an attack that we call sponge poisoning. This attack allows one to increase the energy consumption and latency of machine-learning models indiscriminately on each test-time input. We present a novel formalization for sponge poisoning, overcoming the limitations related to the optimization of test-time sponge examples, and show that this attack is possible even if the attacker only controls a few model updates; for instance, if model training is outsourced to an untrusted third-party or distributed via federated learning. Our extensive experimental analysis shows that sponge poisoning can almost completely vanish the effect of hardware accelerators. We also analyze the activations of poisoned models, identifying which components are more vulnerable to this attack. Finally, we examine the feasibility of countermeasures against sponge poisoning to decrease energy consumption, showing that sanitization methods may be overly expensive for most of the users.
comment: Preprint;16 pages
♻ ☆ Machine Learning Security against Data Poisoning: Are We There Yet?
The recent success of machine learning (ML) has been fueled by the increasing availability of computing power and large amounts of data in many different applications. However, the trustworthiness of the resulting models can be compromised when such data is maliciously manipulated to mislead the learning process. In this article, we first review poisoning attacks that compromise the training data used to learn ML models, including attacks that aim to reduce the overall performance, manipulate the predictions on specific test samples, and even implant backdoors in the model. We then discuss how to mitigate these attacks using basic security principles, or by deploying ML-oriented defensive mechanisms. We conclude our article by formulating some relevant open challenges which are hindering the development of testing methods and benchmarks suitable for assessing and improving the trustworthiness of ML models against data poisoning attacks
comment: preprint, 10 pages, 3 figures
♻ ☆ Inversion-Based Style Transfer with Diffusion Models CVPR 2023
The artistic style within a painting is the means of expression, which includes not only the painting material, colors, and brushstrokes, but also the high-level attributes including semantic elements, object shapes, etc. Previous arbitrary example-guided artistic image generation methods often fail to control shape changes or convey elements. The pre-trained text-to-image synthesis diffusion probabilistic models have achieved remarkable quality, but it often requires extensive textual descriptions to accurately portray attributes of a particular painting. We believe that the uniqueness of an artwork lies precisely in the fact that it cannot be adequately explained with normal language. Our key idea is to learn artistic style directly from a single painting and then guide the synthesis without providing complex textual descriptions. Specifically, we assume style as a learnable textual description of a painting. We propose an inversion-based style transfer method (InST), which can efficiently and accurately learn the key information of an image, thus capturing and transferring the artistic style of a painting. We demonstrate the quality and efficiency of our method on numerous paintings of various artists and styles. Code and models are available at https://github.com/zyxElsa/InST.
comment: accepted by CVPR 2023
♻ ☆ X-Avatar: Expressive Human Avatars
We present X-Avatar, a novel avatar model that captures the full expressiveness of digital humans to bring about life-like experiences in telepresence, AR/VR and beyond. Our method models bodies, hands, facial expressions and appearance in a holistic fashion and can be learned from either full 3D scans or RGB-D data. To achieve this, we propose a part-aware learned forward skinning module that can be driven by the parameter space of SMPL-X, allowing for expressive animation of X-Avatars. To efficiently learn the neural shape and deformation fields, we propose novel part-aware sampling and initialization strategies. This leads to higher fidelity results, especially for smaller body parts while maintaining efficient training despite increased number of articulated bones. To capture the appearance of the avatar with high-frequency details, we extend the geometry and deformation fields with a texture network that is conditioned on pose, facial expression, geometry and the normals of the deformed surface. We show experimentally that our method outperforms strong baselines in both data domains both quantitatively and qualitatively on the animation task. To facilitate future research on expressive avatars we contribute a new dataset, called X-Humans, containing 233 sequences of high-quality textured scans from 20 participants, totalling 35,500 data frames.
comment: Project page: https://skype-line.github.io/projects/X-Avatar/
♻ ☆ Learning Trajectory-Word Alignments for Video-Language Tasks
In a video, an object usually appears as the trajectory, i.e., it spans over a few spatial but longer temporal patches, that contains abundant spatiotemporal contexts. However, modern Video-Language BERTs (VDL-BERTs) neglect this trajectory characteristic that they usually follow image-language BERTs (IL-BERTs) to deploy the patch-to-word (P2W) attention that may over-exploit trivial spatial contexts and neglect significant temporal contexts. To amend this, we propose a novel TW-BERT to learn Trajectory-Word alignment by a newly designed trajectory-to-word (T2W) attention for solving video-language tasks. Moreover, previous VDL-BERTs usually uniformly sample a few frames into the model while different trajectories have diverse graininess, i.e., some trajectories span longer frames and some span shorter, and using a few frames will lose certain useful temporal contexts. However, simply sampling more frames will also make pre-training infeasible due to the largely increased training burdens. To alleviate the problem, during the fine-tuning stage, we insert a novel Hierarchical Frame-Selector (HFS) module into the video encoder. HFS gradually selects the suitable frames conditioned on the text context for the later cross-modal encoder to learn better trajectory-word alignments. By the proposed T2W attention and HFS, our TW-BERT achieves SOTA performances on text-to-video retrieval tasks, and comparable performances on video question-answering tasks with some VDL-BERTs trained on much more data. The code will be available in the supplementary material.
♻ ☆ MobileBrick: Building LEGO for 3D Reconstruction on Mobile Devices CVPR 2023
High-quality 3D ground-truth shapes are critical for 3D object reconstruction evaluation. However, it is difficult to create a replica of an object in reality, and even 3D reconstructions generated by 3D scanners have artefacts that cause biases in evaluation. To address this issue, we introduce a novel multi-view RGBD dataset captured using a mobile device, which includes highly precise 3D ground-truth annotations for 153 object models featuring a diverse set of 3D structures. We obtain precise 3D ground-truth shape without relying on high-end 3D scanners by utilising LEGO models with known geometry as the 3D structures for image capture. The distinct data modality offered by high-resolution RGB images and low-resolution depth maps captured on a mobile device, when combined with precise 3D geometry annotations, presents a unique opportunity for future research on high-fidelity 3D reconstruction. Furthermore, we evaluate a range of 3D reconstruction algorithms on the proposed dataset. Project page: http://code.active.vision/MobileBrick/
comment: To be appeared at CVPR 2023
♻ ☆ ConZIC: Controllable Zero-shot Image Captioning by Sampling-Based Polishing CVPR2023
Zero-shot capability has been considered as a new revolution of deep learning, letting machines work on tasks without curated training data. As a good start and the only existing outcome of zero-shot image captioning (IC), ZeroCap abandons supervised training and sequentially searches every word in the caption using the knowledge of large-scale pretrained models. Though effective, its autoregressive generation and gradient-directed searching mechanism limit the diversity of captions and inference speed, respectively. Moreover, ZeroCap does not consider the controllability issue of zero-shot IC. To move forward, we propose a framework for Controllable Zero-shot IC, named ConZIC. The core of ConZIC is a novel sampling-based non-autoregressive language model named GibbsBERT, which can generate and continuously polish every word. Extensive quantitative and qualitative results demonstrate the superior performance of our proposed ConZIC for both zero-shot IC and controllable zero-shot IC. Especially, ConZIC achieves about 5x faster generation speed than ZeroCap, and about 1.5x higher diversity scores, with accurate generation given different control signals.
comment: Accepted by CVPR2023
♻ ☆ MedKLIP: Medical Knowledge Enhanced Language-Image Pre-Training
In this paper, we consider enhancing medical visual-language pre-training (VLP) with domain-specific knowledge, by exploiting the paired image-text reports from the radiological daily practice. In particular, we make the following contributions: First, unlike existing works that directly process the raw reports, we adopt a novel triplet extraction module to extract the medical-related information, avoiding unnecessary complexity from language grammar and enhancing the supervision signals; Second, we propose a novel triplet encoding module with entity translation by querying a knowledge base, to exploit the rich domain knowledge in medical field, and implicitly build relationships between medical entities in the language embedding space; Third, we propose to use a Transformer-based fusion model for spatially aligning the entity description with visual signals at the image patch level, enabling the ability for medical diagnosis; Fourth, we conduct thorough experiments to validate the effectiveness of our architecture, and benchmark on numerous public benchmarks, e.g., ChestX-ray14, RSNA Pneumonia, SIIM-ACR Pneumothorax, COVIDx CXR-2, COVID Rural, and EdemaSeverity. In both zero-shot and fine-tuning settings, our model has demonstrated strong performance compared with the former methods on disease classification and grounding.
♻ ☆ Self-Supervised Learning for Place Representation Generalization across Appearance Changes
Visual place recognition is a key to unlocking spatial navigation for animals, humans and robots. While state-of-the-art approaches are trained in a supervised manner and therefore hardly capture the information needed for generalizing to unusual conditions, we argue that self-supervised learning may help abstracting the place representation so that it can be foreseen, irrespective of the conditions. More precisely, in this paper, we investigate learning features that are robust to appearance modifications while sensitive to geometric transformations in a self-supervised manner. This dual-purpose training is made possible by combining the two self-supervision main paradigms, \textit{i.e.} contrastive and predictive learning. Our results on standard benchmarks reveal that jointly learning such appearance-robust and geometry-sensitive image descriptors leads to competitive visual place recognition results across adverse seasonal and illumination conditions, without requiring any human-annotated labels.
comment: 11 pages, 6 figures
♻ ☆ Uncertainty-Aware DNN for Multi-Modal Camera Localization
Camera localization, i.e., camera pose regression, represents an important task in computer vision since it has many practical applications such as in the context of intelligent vehicles and their localization. Having reliable estimates of the regression uncertainties is also important, as it would allow us to catch dangerous localization failures. In the literature, uncertainty estimation in Deep Neural Networks (DNNs) is often performed through sampling methods, such as Monte Carlo Dropout (MCD) and Deep Ensemble (DE), at the expense of undesirable execution time or an increase in hardware resources. In this work, we considered an uncertainty estimation approach named Deep Evidential Regression (DER) that avoids any sampling technique, providing direct uncertainty estimates. Our goal is to provide a systematic approach to intercept localization failures of camera localization systems based on DNNs architectures, by analyzing the generated uncertainties. We propose to exploit CMRNet, a DNN approach for multi-modal image to LiDAR map registration, by modifying its internal configuration to allow for extensive experimental activity on the KITTI dataset. The experimental section highlights CMRNet's major flaws and proves that our proposal does not compromise the original localization performances but also provides, at the same time, the necessary introspection measures that would allow end-users to act accordingly.
♻ ☆ TetraSphere: A Neural Descriptor for O(3)-Invariant Point Cloud Classification
Rotation invariance is an important requirement for the analysis of 3D point clouds. In this paper, we present a learnable descriptor for rotation- and reflection-invariant 3D point cloud classification based on recently introduced steerable 3D spherical neurons and vector neurons. Specifically, we show that the two approaches are compatible, and we show how to apply steerable neurons in an end-to-end method for the first time. In our approach, we perform TetraTransform -- which lifts the 3D input to an equivariant 4D representation, constructed by the steerable neurons -- and extract deeper rotation-equivariant features using vector neurons, subsequently computing pair-wise O(3)-invariant inner products of these features. This integration of the TetraTransform into the VN-DGCNN framework, termed TetraSphere, is used to classify synthetic and real-world data in arbitrary orientations. Taking only 3D coordinates as input, TetraSphere sets a new state-of-the-art classification performance on randomly rotated objects of the hardest subset of ScanObjectNN, even when trained on data without additional rotation augmentation. Our results reveal the practical value of spherical decision surfaces for learning in 3D Euclidean space.
♻ ☆ Patch of Invisibility: Naturalistic Black-Box Adversarial Attacks on Object Detectors
Adversarial attacks on deep-learning models have been receiving increased attention in recent years. Work in this area has mostly focused on gradient-based techniques, so-called white-box attacks, wherein the attacker has access to the targeted model's internal parameters; such an assumption is usually unrealistic in the real world. Some attacks additionally use the entire pixel space to fool a given model, which is neither practical nor physical (i.e., real-world). On the contrary, we propose herein a gradient-free method that uses the learned image manifold of a pretrained generative adversarial network (GAN) to generate naturalistic physical adversarial patches for object detectors. We show that our proposed method works both digitally and physically.
♻ ☆ Anomaly Detection via Multi-Scale Contrasted Memory
Deep anomaly detection (AD) aims to provide robust and efficient classifiers for one-class and unbalanced settings. However current AD models still struggle on edge-case normal samples and are often unable to keep high performance over different scales of anomalies. Moreover, there currently does not exist a unified framework efficiently covering both one-class and unbalanced learnings. In the light of these limitations, we introduce a new two-stage anomaly detector which memorizes during training multi-scale normal prototypes to compute an anomaly deviation score. First, we simultaneously learn representations and memory modules on multiple scales using a novel memory-augmented contrastive learning. Then, we train an anomaly distance detector on the spatial deviation maps between prototypes and observations. Our model highly improves the state-of-the-art performance on a wide range of object, style and local anomalies with up to 50% error relative improvement on CIFAR-100. It is also the first model to keep high performance across the one-class and unbalanced settings.
♻ ☆ ViLPAct: A Benchmark for Compositional Generalization on Multimodal Human Activities EACL2023
We introduce ViLPAct, a novel vision-language benchmark for human activity planning. It is designed for a task where embodied AI agents can reason and forecast future actions of humans based on video clips about their initial activities and intents in text. The dataset consists of 2.9k videos from \charades extended with intents via crowdsourcing, a multi-choice question test set, and four strong baselines. One of the baselines implements a neurosymbolic approach based on a multi-modal knowledge base (MKB), while the other ones are deep generative models adapted from recent state-of-the-art (SOTA) methods. According to our extensive experiments, the key challenges are compositional generalization and effective use of information from both modalities.
comment: Accepted at EACL2023 (Findings)
♻ ☆ 3D Former: Monocular Scene Reconstruction with 3D SDF Transformers ICLR 2023
Monocular scene reconstruction from posed images is challenging due to the complexity of a large environment. Recent volumetric methods learn to directly predict the TSDF volume and have demonstrated promising results in this task. However, most methods focus on how to extract and fuse the 2D features to a 3D feature volume, but none of them improve the way how the 3D volume is aggregated. In this work, we propose an SDF transformer network, which replaces the role of 3D CNN for better 3D feature aggregation. To reduce the explosive computation complexity of the 3D multi-head attention, we propose a sparse window attention module, where the attention is only calculated between the non-empty voxels within a local window. Then a top-down-bottom-up 3D attention network is built for 3D feature aggregation, where a dilate-attention structure is proposed to prevent geometry degeneration, and two global modules are employed to equip with global receptive fields. The experiments on multiple datasets show that this 3D transformer network generates a more accurate and complete reconstruction, which outperforms previous methods by a large margin. Remarkably, the mesh accuracy is improved by 41.8%, and the mesh completeness is improved by 25.3% on the ScanNet dataset. Project page: https://weihaosky.github.io/sdfformer.
comment: Accepted to ICLR 2023
♻ ☆ Learning Spatial-Frequency Transformer for Visual Object Tracking
Recent trackers adopt the Transformer to combine or replace the widely used ResNet as their new backbone network. Although their trackers work well in regular scenarios, however, they simply flatten the 2D features into a sequence to better match the Transformer. We believe these operations ignore the spatial prior of the target object which may lead to sub-optimal results only. In addition, many works demonstrate that self-attention is actually a low-pass filter, which is independent of input features or key/queries. That is to say, it may suppress the high-frequency component of the input features and preserve or even amplify the low-frequency information. To handle these issues, in this paper, we propose a unified Spatial-Frequency Transformer that models the Gaussian spatial Prior and High-frequency emphasis Attention (GPHA) simultaneously. To be specific, Gaussian spatial prior is generated using dual Multi-Layer Perceptrons (MLPs) and injected into the similarity matrix produced by multiplying Query and Key features in self-attention. The output will be fed into a Softmax layer and then decomposed into two components, i.e., the direct signal and high-frequency signal. The low- and high-pass branches are rescaled and combined to achieve all-pass, therefore, the high-frequency features will be protected well in stacked self-attention layers. We further integrate the Spatial-Frequency Transformer into the Siamese tracking framework and propose a novel tracking algorithm, termed SFTransT. The cross-scale fusion based SwinTransformer is adopted as the backbone, and also a multi-head cross-attention module is used to boost the interaction between search and template features. The output will be fed into the tracking head for target localization. Extensive experiments on both short-term and long-term tracking benchmarks all demonstrate the effectiveness of our proposed framework.
♻ ☆ Rethinking the Objectives of Vector-Quantized Tokenizers for Image Synthesis
Vector-Quantized (VQ-based) generative models usually consist of two basic components, i.e., VQ tokenizers and generative transformers. Prior research focuses on improving the reconstruction fidelity of VQ tokenizers but rarely examines how the improvement in reconstruction affects the generation ability of generative transformers. In this paper, we surprisingly find that improving the reconstruction fidelity of VQ tokenizers does not necessarily improve the generation. Instead, learning to compress semantic features within VQ tokenizers significantly improves generative transformers' ability to capture textures and structures. We thus highlight two competing objectives of VQ tokenizers for image synthesis: semantic compression and details preservation. Different from previous work that only pursues better details preservation, we propose Semantic-Quantized GAN (SeQ-GAN) with two learning phases to balance the two objectives. In the first phase, we propose a semantic-enhanced perceptual loss for better semantic compression. In the second phase, we fix the encoder and codebook, but enhance and finetune the decoder to achieve better details preservation. The proposed SeQ-GAN greatly improves VQ-based generative models and surpasses the GAN and Diffusion Models on both unconditional and conditional image generation. Our SeQ-GAN (364M) achieves Frechet Inception Distance (FID) of 6.25 and Inception Score (IS) of 140.9 on 256x256 ImageNet generation, a remarkable improvement over VIT-VQGAN (714M), which obtains 11.2 FID and 97.2 IS.
♻ ☆ Aerial View Localization with Reinforcement Learning: Towards Emulating Search-and-Rescue ICLR 2023
Climate-induced disasters are and will continue to be on the rise, and thus search-and-rescue (SAR) operations, where the task is to localize and assist one or several people who are missing, become increasingly relevant. In many cases the rough location may be known and a UAV can be deployed to explore a given, confined area to precisely localize the missing people. Due to time and battery constraints it is often critical that localization is performed as efficiently as possible. In this work we approach this type of problem by abstracting it as an aerial view goal localization task in a framework that emulates a SAR-like setup without requiring access to actual UAVs. In this framework, an agent operates on top of an aerial image (proxy for a search area) and is tasked with localizing a goal that is described in terms of visual cues. To further mimic the situation on an actual UAV, the agent is not able to observe the search area in its entirety, not even at low resolution, and thus it has to operate solely based on partial glimpses when navigating towards the goal. To tackle this task, we propose AiRLoc, a reinforcement learning (RL)-based model that decouples exploration (searching for distant goals) and exploitation (localizing nearby goals). Extensive evaluations show that AiRLoc outperforms heuristic search methods as well as alternative learnable approaches, and that it generalizes across datasets, e.g. to disaster-hit areas without seeing a single disaster scenario during training. We also conduct a proof-of-concept study which indicates that the learnable methods outperform humans on average. Code and models have been made publicly available at https://github.com/aleksispi/airloc.
comment: Accepted to ICLR 2023 Workshop on Machine Learning for Remote Sensing
♻ ☆ 6IMPOSE: Bridging the Reality Gap in 6D Pose Estimation for Robotic Grasping
6D pose recognition has been a crucial factor in the success of robotic grasping, and recent deep learning based approaches have achieved remarkable results on benchmarks. However, their generalization capabilities in real-world applications remain unclear. To overcome this gap, we introduce 6IMPOSE, a novel framework for sim-to-real data generation and 6D pose estimation. 6IMPOSE consists of four modules: First, a data generation pipeline that employs the 3D software suite Blender to create synthetic RGBD image datasets with 6D pose annotations. Second, an annotated RGBD dataset of five household objects generated using the proposed pipeline. Third, a real-time two-stage 6D pose estimation approach that integrates the object detector YOLO-V4 and a streamlined, real-time version of the 6D pose estimation algorithm PVN3D optimized for time-sensitive robotics applications. Fourth, a codebase designed to facilitate the integration of the vision system into a robotic grasping experiment. Our approach demonstrates the efficient generation of large amounts of photo-realistic RGBD images and the successful transfer of the trained inference model to robotic grasping experiments, achieving an overall success rate of 87% in grasping five different household objects from cluttered backgrounds under varying lighting conditions. This is made possible by the fine-tuning of data generation and domain randomization techniques, and the optimization of the inference pipeline, overcoming the generalization and performance shortcomings of the original PVN3D algorithm. Finally, we make the code, synthetic dataset, and all the pretrained models available on Github.
♻ ☆ Towards Good Practices in Evaluating Transfer Adversarial Attacks
Transfer adversarial attacks raise critical security concerns in real-world, black-box scenarios. However, the actual progress of this field is difficult to assess due to two common limitations in existing evaluations. First, different methods are often not systematically and fairly evaluated in a one-to-one comparison. Second, only transferability is evaluated but another key attack property, stealthiness, is largely overlooked. In this work, we design good practices to address these limitations, and we present the first comprehensive evaluation of transfer attacks, covering 23 representative attacks against 9 defenses on ImageNet. In particular, we propose to categorize existing attacks into five categories, which enables our systematic category-wise analyses. These analyses lead to new findings that even challenge existing knowledge and also help determine the optimal attack hyperparameters for our attack-wise comprehensive evaluation. We also pay particular attention to stealthiness, by adopting diverse imperceptibility metrics and looking into new, finer-grained characteristics. Overall, our new insights into transferability and stealthiness lead to actionable good practices for future evaluations.
comment: Our code and a list of categorized attacks are publicly available at https://github.com/ZhengyuZhao/TransferAttackEval
♻ ☆ Hair and Scalp Disease Detection using Machine Learning and Image Processing
Almost 80 million Americans suffer from hair loss due to aging, stress, medication, or genetic makeup. Hair and scalp-related diseases often go unnoticed in the beginning. Sometimes, a patient cannot differentiate between hair loss and regular hair fall. Diagnosing hair-related diseases is time-consuming as it requires professional dermatologists to perform visual and medical tests. Because of that, the overall diagnosis gets delayed, which worsens the severity of the illness. Due to the image-processing ability, neural network-based applications are used in various sectors, especially healthcare and health informatics, to predict deadly diseases like cancers and tumors. These applications assist clinicians and patients and provide an initial insight into early-stage symptoms. In this study, we used a deep learning approach that successfully predicts three main types of hair loss and scalp-related diseases: alopecia, psoriasis, and folliculitis. However, limited study in this area, unavailability of a proper dataset, and degree of variety among the images scattered over the internet made the task challenging. 150 images were obtained from various sources and then preprocessed by denoising, image equalization, enhancement, and data balancing, thereby minimizing the error rate. After feeding the processed data into the 2D convolutional neural network (CNN) model, we obtained overall training accuracy of 96.2%, with a validation accuracy of 91.1%. The precision and recall score of alopecia, psoriasis, and folliculitis are 0.895, 0.846, and 1.0, respectively. We also created a dataset of the scalp images for future prospective researchers.
♻ ☆ Fourier Series Expansion Based Filter Parametrization for Equivariant Convolutions
It has been shown that equivariant convolution is very helpful for many types of computer vision tasks. Recently, the 2D filter parametrization technique plays an important role when designing equivariant convolutions. However, the current filter parametrization method still has its evident drawbacks, where the most critical one lies in the accuracy problem of filter representation. Against this issue, in this paper we modify the classical Fourier series expansion for 2D filters, and propose a new set of atomic basis functions for filter parametrization. The proposed filter parametrization method not only finely represents 2D filters with zero error when the filter is not rotated, but also substantially alleviates the fence-effect-caused quality degradation when the filter is rotated. Accordingly, we construct a new equivariant convolution method based on the proposed filter parametrization method, named F-Conv. We prove that the equivariance of the proposed F-Conv is exact in the continuous domain, which becomes approximate only after discretization. Extensive experiments show the superiority of the proposed method. Particularly, we adopt rotation equivariant convolution methods to image super-resolution task, and F-Conv evidently outperforms previous filter parametrization based method in this task, reflecting its intrinsic capability of faithfully preserving rotation symmetries in local image features.
comment: 15 pages, 8 figures
♻ ☆ Fewer is More: Efficient Object Detection in Large Aerial Images
Current mainstream object detection methods for large aerial images usually divide large images into patches and then exhaustively detect the objects of interest on all patches, no matter whether there exist objects or not. This paradigm, although effective, is inefficient because the detectors have to go through all patches, severely hindering the inference speed. This paper presents an Objectness Activation Network (OAN) to help detectors focus on fewer patches but achieve more efficient inference and more accurate results, enabling a simple and effective solution to object detection in large images. In brief, OAN is a light fully-convolutional network for judging whether each patch contains objects or not, which can be easily integrated into many object detectors and jointly trained with them end-to-end. We extensively evaluate our OAN with five advanced detectors. Using OAN, all five detectors acquire more than 30.0% speed-up on three large-scale aerial image datasets, meanwhile with consistent accuracy improvements. On extremely large Gaofen-2 images (29200$\times$27620 pixels), our OAN improves the detection speed by 70.5%. Moreover, we extend our OAN to driving-scene object detection and 4K video object detection, boosting the detection speed by 112.1% and 75.0%, respectively, without sacrificing the accuracy. Code is available at https://github.com/Ranchosky/OAN.
comment: This manuscript is the accepted version for SCIENCE CHINA Information Sciences
♻ ☆ Uncertainty-Aware Unsupervised Image Deblurring with Deep Residual Prior
Non-blind deblurring methods achieve decent performance under the accurate blur kernel assumption. Since the kernel uncertainty (i.e. kernel error) is inevitable in practice, semi-blind deblurring is suggested to handle it by introducing the prior of the kernel (or induced) error. However, how to design a suitable prior for the kernel (or induced) error remains challenging. Hand-crafted prior, incorporating domain knowledge, generally performs well but may lead to poor performance when kernel (or induced) error is complex. Data-driven prior, which excessively depends on the diversity and abundance of training data, is vulnerable to out-of-distribution blurs and images. To address this challenge, we suggest a dataset-free deep residual prior for the kernel induced error (termed as residual) expressed by a customized untrained deep neural network, which allows us to flexibly adapt to different blurs and images in real scenarios. By organically integrating the respective strengths of deep priors and hand-crafted priors, we propose an unsupervised semi-blind deblurring model which recovers the latent image from the blurry image and inaccurate blur kernel. To tackle the formulated model, an efficient alternating minimization algorithm is developed. Extensive experiments demonstrate the favorable performance of the proposed method as compared to data-driven and model-driven methods in terms of image quality and the robustness to the kernel error.
♻ ☆ Exploring Deep Models for Practical Gait Recognition
Gait recognition is a rapidly advancing vision technique for person identification from a distance. Prior studies predominantly employed relatively small and shallow neural networks to extract subtle gait features, achieving impressive successes in indoor settings. Nevertheless, experiments revealed that these existing methods mostly produce unsatisfactory results when applied to newly released in-the-wild gait datasets. This paper presents a unified perspective to explore how to construct deep models for state-of-the-art outdoor gait recognition, including the classical CNN-based and emerging Transformer-based architectures. Consequently, we emphasize the importance of suitable network capacity, explicit temporal modeling, and deep transformer structure for discriminative gait representation learning. Our proposed CNN-based DeepGaitV2 series and Transformer-based SwinGait series exhibit significant performance gains in outdoor scenarios, \textit{e.g.}, about +30\% rank-1 accuracy compared with many state-of-the-art methods on the challenging GREW dataset. This work is expected to further boost the research and application of gait recognition. Code will be available at https://github.com/ShiqiYu/OpenGait.
♻ ☆ Decoupling Skill Learning from Robotic Control for Generalizable Object Manipulation ICRA
Recent works in robotic manipulation through reinforcement learning (RL) or imitation learning (IL) have shown potential for tackling a range of tasks e.g., opening a drawer or a cupboard. However, these techniques generalize poorly to unseen objects. We conjecture that this is due to the high-dimensional action space for joint control. In this paper, we take an alternative approach and separate the task of learning 'what to do' from 'how to do it' i.e., whole-body control. We pose the RL problem as one of determining the skill dynamics for a disembodied virtual manipulator interacting with articulated objects. The whole-body robotic kinematic control is optimized to execute the high-dimensional joint motion to reach the goals in the workspace. It does so by solving a quadratic programming (QP) model with robotic singularity and kinematic constraints. Our experiments on manipulating complex articulated objects show that the proposed approach is more generalizable to unseen objects with large intra-class variations, outperforming previous approaches. The evaluation results indicate that our approach generates more compliant robotic motion and outperforms the pure RL and IL baselines in task success rates. Additional information and videos are available at https://kl-research.github.io/decoupskill
comment: Accepted to IEEE International Conference on Robotics and Automation (ICRA) 2023
♻ ☆ Hypergraph Transformer for Skeleton-based Action Recognition
Skeleton-based action recognition aims to predict human actions given human joint coordinates with skeletal interconnections. To model such off-grid data points and their co-occurrences, Transformer-based formulations would be a natural choice. However, Transformers still lag behind state-of-the-art methods using graph convolutional networks (GCNs). Transformers assume that the input is permutation-invariant and homogeneous (partially alleviated by positional encoding), which ignores an important characteristic of skeleton data, i.e., bone connectivity. Furthermore, each type of body joint has a clear physical meaning in human motion, i.e., motion retains an intrinsic relationship regardless of the joint coordinates, which is not explored in Transformers. In fact, certain re-occurring groups of body joints are often involved in specific actions, such as the subconscious hand movement for keeping balance. Vanilla attention is incapable of describing such underlying relations that are persistent and beyond pair-wise. In this work, we aim to exploit these unique aspects of skeleton data to close the performance gap between Transformers and GCNs. Specifically, we propose a new self-attention (SA) extension, named Hypergraph Self-Attention (HyperSA), to incorporate inherently higher-order relations into the model. The K-hop relative positional embeddings are also employed to take bone connectivity into account. We name the resulting model Hyperformer, and it achieves comparable or better performance w.r.t. accuracy and efficiency than state-of-the-art GCN architectures on NTU RGB+D, NTU RGB+D 120, and Northwestern-UCLA datasets. On the largest NTU RGB+D 120 dataset, the significantly improved performance reached by our Hyperformer demonstrates the underestimated potential of Transformer models in this field.
♻ ☆ Evaluation and Improvement of Interpretability for Self-Explainable Part-Prototype Networks
Part-prototype networks (e.g., ProtoPNet, ProtoTree and ProtoPool) have attracted broad research interest for their intrinsic interpretability and comparable accuracy to non-interpretable counterparts. However, recent works find that the interpretability from prototypes is fragile, due to the semantic gap between the similarities in the feature space and that in the input space. In this work, we strive to address this challenge by making the first attempt to quantitatively and objectively evaluate the interpretability of the part-prototype networks. Specifically, we propose two evaluation metrics, termed as consistency score and stability score, to evaluate the explanation consistency across images and the explanation robustness against perturbations, respectively, both of which are essential for explanations taken into practice. Furthermore, we propose an elaborated part-prototype network with a shallow-deep feature alignment (SDFA) module and a score aggregation (SA) module to improve the interpretability of prototypes. We conduct systematical evaluation experiments and provide substantial discussions to uncover the interpretability of existing part-prototype networks. Experiments on three benchmarks across nine architectures demonstrate that our model achieves significantly superior performance to the state of the art, in both the accuracy and interpretability. Codes are available at https://github.com/hqhQAQ/EvalProtoPNet.
♻ ☆ UKnow: A Unified Knowledge Protocol for Common-Sense Reasoning and Vision-Language Pre-training
This work presents a unified knowledge protocol, called UKnow, which facilitates knowledge-based studies from the perspective of data. Particularly focusing on visual and linguistic modalities, we categorize data knowledge into five unit types, namely, in-image, in-text, cross-image, cross-text, and image-text, and set up an efficient pipeline to help construct the multimodal knowledge graph from any data collection. Thanks to the logical information naturally contained in knowledge graph, organizing datasets under UKnow format opens up more possibilities of data usage compared to the commonly used image-text pairs. Following UKnow protocol, we collect, from public international news, a large-scale multimodal knowledge graph dataset that consists of 1,388,568 nodes (with 571,791 vision-related ones) and 3,673,817 triplets. The dataset is also annotated with rich event tags, including 11 coarse labels and 9,185 fine labels. Experiments on four benchmarks demonstrate the potential of UKnow in supporting common-sense reasoning and boosting vision-language pre-training with a single dataset, benefiting from its unified form of knowledge organization. Code, dataset, and models will be made publicly available.
♻ ☆ Flat Multi-modal Interaction Transformer for Named Entity Recognition COLING 2022
Multi-modal named entity recognition (MNER) aims at identifying entity spans and recognizing their categories in social media posts with the aid of images. However, in dominant MNER approaches, the interaction of different modalities is usually carried out through the alternation of self-attention and cross-attention or over-reliance on the gating machine, which results in imprecise and biased correspondence between fine-grained semantic units of text and image. To address this issue, we propose a Flat Multi-modal Interaction Transformer (FMIT) for MNER. Specifically, we first utilize noun phrases in sentences and general domain words to obtain visual cues. Then, we transform the fine-grained semantic representation of the vision and text into a unified lattice structure and design a novel relative position encoding to match different modalities in Transformer. Meanwhile, we propose to leverage entity boundary detection as an auxiliary task to alleviate visual bias. Experiments show that our methods achieve the new state-of-the-art performance on two benchmark datasets.
comment: Accepted by COLING 2022, oral paper
♻ ☆ Indiscriminate Poisoning Attacks on Unsupervised Contrastive Learning ICLR 2023
Indiscriminate data poisoning attacks are quite effective against supervised learning. However, not much is known about their impact on unsupervised contrastive learning (CL). This paper is the first to consider indiscriminate poisoning attacks of contrastive learning. We propose Contrastive Poisoning (CP), the first effective such attack on CL. We empirically show that Contrastive Poisoning, not only drastically reduces the performance of CL algorithms, but also attacks supervised learning models, making it the most generalizable indiscriminate poisoning attack. We also show that CL algorithms with a momentum encoder are more robust to indiscriminate poisoning, and propose a new countermeasure based on matrix completion. Code is available at: https://github.com/kaiwenzha/contrastive-poisoning.
comment: ICLR 2023 Spotlight (notable top 25%). The first two authors contributed equally to this paper
♻ ☆ Aggregation of Disentanglement: Reconsidering Domain Variations in Domain Generalization
Domain Generalization (DG) is a fundamental challenge for machine learning models, which aims to improve model generalization on various domains. Previous methods focus on generating domain invariant features from various source domains. However, we argue that the domain variantions also contain useful information, ie, classification-aware information, for downstream tasks, which has been largely ignored. Different from learning domain invariant features from source domains, we decouple the input images into Domain Expert Features and noise. The proposed domain expert features lie in a learned latent space where the images in each domain can be classified independently, enabling the implicit use of classification-aware domain variations. Based on the analysis, we proposed a novel paradigm called Domain Disentanglement Network (DDN) to disentangle the domain expert features from the source domain images and aggregate the source domain expert features for representing the target test domain. We also propound a new contrastive learning method to guide the domain expert features to form a more balanced and separable feature space. Experiments on the widely-used benchmarks of PACS, VLCS, OfficeHome, DomainNet, and TerraIncognita demonstrate the competitive performance of our method compared to the recently proposed alternatives.
♻ ☆ DP-NeRF: Deblurred Neural Radiance Field with Physical Scene Priors CVPR 2023
Neural Radiance Field (NeRF) has exhibited outstanding three-dimensional (3D) reconstruction quality via the novel view synthesis from multi-view images and paired calibrated camera parameters. However, previous NeRF-based systems have been demonstrated under strictly controlled settings, with little attention paid to less ideal scenarios, including with the presence of noise such as exposure, illumination changes, and blur. In particular, though blur frequently occurs in real situations, NeRF that can handle blurred images has received little attention. The few studies that have investigated NeRF for blurred images have not considered geometric and appearance consistency in 3D space, which is one of the most important factors in 3D reconstruction. This leads to inconsistency and the degradation of the perceptual quality of the constructed scene. Hence, this paper proposes a DP-NeRF, a novel clean NeRF framework for blurred images, which is constrained with two physical priors. These priors are derived from the actual blurring process during image acquisition by the camera. DP-NeRF proposes rigid blurring kernel to impose 3D consistency utilizing the physical priors and adaptive weight proposal to refine the color composition error in consideration of the relationship between depth and blur. We present extensive experimental results for synthetic and real scenes with two types of blur: camera motion blur and defocus blur. The results demonstrate that DP-NeRF successfully improves the perceptual quality of the constructed NeRF ensuring 3D geometric and appearance consistency. We further demonstrate the effectiveness of our model with comprehensive ablation analysis.
comment: Accepted at CVPR 2023, Code: https://github.com/dogyoonlee/DP-NeRF, Project page: https://dogyoonlee.github.io/dpnerf/
♻ ☆ Towards Connectivity-Aware Pulmonary Airway Segmentation
Detailed pulmonary airway segmentation is a clinically important task for endobronchial intervention and treatment of peripheral pulmonary lesions. Convolutional Neural Networks (CNNs) are promising for automated analysis of medical imaging, which however performs poorly on airway segmentation. Specifically, breakage of small bronchi distals cannot be effectively eliminated in the prediction results of CNNs, which is detrimental to use as a reference for bronchoscopic-assisted surgery. In this paper, we proposed a connectivity-aware segmentation framework to improve the performance of airway segmentation. A Connectivity-Aware Surrogate (CAS) module is first proposed to balance the training progress within-class distribution. Furthermore, a Local-Sensitive Distance (LSD) module is designed to identify the breakage and minimize the variation of the distance map between the prediction and ground-truth. The proposed method is validated with the publically available reference airway segmentation datasets. The detected rate of branch and length on public EXACT'09 and BAS datasets are 82.1%/79.6% and 96.5%/91.5% respectively, demonstrating the effectiveness of the method in terms of improving the connectedness of the segmentation performance.
comment: 11 pages, 7 figures
♻ ☆ TORE: Token Reduction for Efficient Human Mesh Recovery with Transformer
In this paper, we introduce a set of simple yet effective TOken REduction (TORE) strategies for Transformer-based Human Mesh Recovery from monocular images. Current SOTA performance is achieved by Transformer-based structures. However, they suffer from high model complexity and computation cost caused by redundant tokens. We propose token reduction strategies based on two important aspects, i.e., the 3D geometry structure and 2D image feature, where we hierarchically recover the mesh geometry with priors from body structure and conduct token clustering to pass fewer but more discriminative image feature tokens to the Transformer. Our method massively reduces the number of tokens involved in high-complexity interactions in the Transformer. This leads to a significantly reduced computational cost while still achieving competitive or even higher accuracy in shape recovery. Extensive experiments across a wide range of benchmarks validate the superior effectiveness of the proposed method. We further demonstrate the generalizability of our method on hand mesh recovery. Our code will be publicly available once the paper is published.
♻ ☆ MonoNeRF: Learning a Generalizable Dynamic Radiance Field from Monocular Videos
In this paper, we target at the problem of learning a generalizable dynamic radiance field from monocular videos. Different from most existing NeRF methods that are based on multiple views, monocular videos only contain one view at each timestamp, thereby suffering from ambiguity along the view direction in estimating point features and scene flows. Previous studies such as DynNeRF disambiguate point features by positional encoding, which is not transferable and severely limits the generalization ability. As a result, these methods have to train one independent model for each scene and suffer from heavy computational costs when applying to increasing monocular videos in real-world applications. To address this, We propose MonoNeRF to simultaneously learn point features and scene flows with point trajectory and feature correspondence constraints across frames. More specifically, we learn an implicit velocity field to estimate point trajectory from temporal features with Neural ODE, which is followed by a flow-based feature aggregation module to obtain spatial features along the point trajectory. We jointly optimize temporal and spatial features in an end-to-end manner. Experiments show that our MonoNeRF is able to learn from multiple scenes and support new applications such as scene editing, unseen frame synthesis, and fast novel scene adaptation.
♻ ☆ RainDiffusion: When Unsupervised Learning Meets Diffusion Models for Real-world Image Deraining
Recent diffusion models show great potential in generative modeling tasks. This motivates us to raise an intriguing question - What will happen when unsupervised learning meets diffusion models for real-world image deraining? Before answering it, we observe two major obstacles of diffusion models in real-world image deraining: the need for paired training data and the limited utilization of multi-scale rain patterns. To overcome the obstacles, we propose RainDiffusion, the first real-world image deraining paradigm based on diffusion models. RainDiffusion is a non-adversarial training paradigm that introduces stable training of unpaired real-world data, rather than weakly adversarial training, serving as a new standard bar for real-world image deraining. It consists of two cooperative branches: Non-diffusive Translation Branch (NTB) and Diffusive Translation Branch (DTB). NTB exploits a cycle-consistent architecture to bypass the difficulty in unpaired training of regular diffusion models by generating initial clean/rainy image pairs. Given initial image pairs, DTB leverages multi-scale diffusion models to progressively refine the desired output via diffusive generative and multi-scale priors, to obtain a better generalization capacity of real-world image deraining. Extensive experiments confirm the superiority of our RainDiffusion over eight un/semi-supervised methods and show its competitive advantages over seven fully-supervised ones.
comment: 9 pages
♻ ☆ Low-Rank Tensor Completion With Generalized CP Decomposition and Nonnegative Integer Tensor Completion
Tensor completion is important to many areas such as computer vision, data analysis, and signal processing. Previously, a category of methods known as low-rank tensor completion has been proposed and developed, involving the enforcement of low-rank structures on completed tensors. While such methods have been constantly improved, none considered exploiting the numerical properties of tensor elements. This work attempts to construct a new methodological framework called GCDTC (Generalized CP Decomposition Tensor Completion) based on numerical properties to achieve higher accuracy in tensor completion. In this newly introduced framework, a generalized form of the CP Decomposition is applied to low-rank tensor completion. This paper also proposes an algorithm known as SPTC (Smooth Poisson Tensor Completion) for nonnegative integer tensor completion as an application of the GCDTC framework. Through experimentation with real-life data, it is verified that this method could produce results superior in completion accuracy to current state-of-the-art methodologies.
comment: 10 pages, 4 figures, and 1 table
♻ ☆ Improving Image Captioning with Control Signal of Sentence Quality ICASSP2023
In the dataset of image captioning, each image is aligned with several descriptions. Despite the fact that the quality of these descriptions varies, existing captioning models treat them equally in the training process. In this paper, we propose a new control signal of sentence quality, which is taken as an additional input to the captioning model. By integrating the control signal information, captioning models are aware of the quality level of the target sentences and handle them differently. Moreover, we propose a novel reinforcement training method specially designed for the control signal of sentence quality: Quality-oriented Self-Annotated Training (Q-SAT). Extensive experiments on MSCOCO dataset show that without extra information from ground truth captions, models controlled by the highest quality level outperform baseline models on accuracy-based evaluation metrics, which validates the effectiveness of our proposed methods.
comment: Accepted by ICASSP2023
♻ ☆ Adaptively Clustering Neighbor Elements for Image Captioning
We design a novel global-local Transformer named \textbf{Ada-ClustFormer} (\textbf{ACF}) to generate captions. We use this name since each layer of ACF can adaptively cluster input elements to carry self-attention (Self-ATT) for learning local context. Compared with other global-local Transformers which carry Self-ATT in fixed-size windows, ACF can capture varying graininess, \eg, an object may cover different numbers of grids or a phrase may contain diverse numbers of words. To build ACF, we insert a probabilistic matrix C into the Self-ATT layer. For an input sequence {{s}_1,...,{s}_N , C_{i,j} softly determines whether the sub-sequence {s_i,...,s_j} should be clustered for carrying Self-ATT. For implementation, {C}_{i,j} is calculated from the contexts of {{s}_i,...,{s}_j}, thus ACF can exploit the input itself to decide which local contexts should be learned. By using ACF to build the vision encoder and language decoder, the captioning model can automatically discover the hidden structures in both vision and language, which encourages the model to learn a unified structural space for transferring more structural commonalities. The experiment results demonstrate the effectiveness of ACF that we achieve CIDEr of 137.8, which outperforms most SOTA captioning models and achieve comparable scores compared with some BERT-based models. The code will be available in the supplementary material.
♻ ☆ CorruptEncoder: Data Poisoning based Backdoor Attacks to Contrastive Learning
Contrastive learning (CL) pre-trains general-purpose encoders using an unlabeled pre-training dataset, which consists of images or image-text pairs. CL is vulnerable to data poisoning based backdoor attacks (DPBAs), in which an attacker injects poisoned inputs into the pre-training dataset so the encoder is backdoored. However, existing DPBAs achieve limited effectiveness. In this work, we propose new DPBAs called CorruptEncoder to CL. CorruptEncoder uses a theory-guided method to create optimal poisoned inputs to maximize attack effectiveness. Our experiments show that CorruptEncoder substantially outperforms existing DPBAs. In particular, CorruptEncoder is the first DPBA that achieves more than 90% attack success rates with only a few (3) reference images and a small poisoning ratio (0.5%). Moreover, we also propose a defense, called localized cropping, to defend against DPBAs. Our results show that our defense can reduce the effectiveness of DPBAs, though it slightly sacrifices the utility of the encoder.
♻ ☆ SFF-DA: Sptialtemporal Feature Fusion for Detecting Anxiety Nonintrusively
Early detection of anxiety is crucial for reducing the suffering of individuals with mental disorders and improving treatment outcomes. Utilizing an mHealth platform for anxiety screening can be particularly practical in improving screening efficiency and reducing costs. However, the effectiveness of existing methods has been hindered by differences in mobile devices used to capture subjects' physical and mental evaluations, as well as by the variability in data quality and small sample size problems encountered in real-world settings. To address these issues, we propose a framework with spatiotemporal feature fusion for detecting anxiety nonintrusively. We use a feature extraction network based on a 3D convolutional network and long short-term memory ("3DCNN+LSTM") to fuse the spatiotemporal features of facial behavior and noncontact physiology, which reduces the impact of uneven data quality. Additionally, we design a similarity assessment strategy to address the issue of deteriorating model accuracy due to small sample sizes. Our framework is validated with a crew dataset from the real world and two public datasets: the University of Burgundy Franche-Comt\'e Psychophysiological (UBFC-Phys) dataset and the Smart Reasoning for Well-being at Home and at Work for Knowledge Work (SWELL-KW) dataset. The experimental results indicate that our framework outperforms the comparison methods.
♻ ☆ Part-Based Models Improve Adversarial Robustness ICLR 2023
We show that combining human prior knowledge with end-to-end learning can improve the robustness of deep neural networks by introducing a part-based model for object classification. We believe that the richer form of annotation helps guide neural networks to learn more robust features without requiring more samples or larger models. Our model combines a part segmentation model with a tiny classifier and is trained end-to-end to simultaneously segment objects into parts and then classify the segmented object. Empirically, our part-based models achieve both higher accuracy and higher adversarial robustness than a ResNet-50 baseline on all three datasets. For instance, the clean accuracy of our part models is up to 15 percentage points higher than the baseline's, given the same level of robustness. Our experiments indicate that these models also reduce texture bias and yield better robustness against common corruptions and spurious correlations. The code is publicly available at https://github.com/chawins/adv-part-model.
comment: Published in ICLR 2023 (poster). Code can be found at https://github.com/chawins/adv-part-model
♻ ☆ Open-Vocabulary Panoptic Segmentation with Text-to-Image Diffusion Models CVPR 2023
We present ODISE: Open-vocabulary DIffusion-based panoptic SEgmentation, which unifies pre-trained text-image diffusion and discriminative models to perform open-vocabulary panoptic segmentation. Text-to-image diffusion models have shown the remarkable capability of generating high-quality images with diverse open-vocabulary language descriptions. This demonstrates that their internal representation space is highly correlated with open concepts in the real world. Text-image discriminative models like CLIP, on the other hand, are good at classifying images into open-vocabulary labels. We propose to leverage the frozen representation of both these models to perform panoptic segmentation of any category in the wild. Our approach outperforms the previous state of the art by significant margins on both open-vocabulary panoptic and semantic segmentation tasks. In particular, with COCO training only, our method achieves 23.4 PQ and 30.0 mIoU on the ADE20K dataset, with 8.3 PQ and 7.9 mIoU absolute improvement over the previous state-of-the-art. Project page is available at https://jerryxu.net/ODISE .
comment: CVPR 2023. Project page: https://jerryxu.net/ODISE
♻ ☆ Imbalanced Open Set Domain Adaptation via Moving-threshold Estimation and Gradual Alignment
Multimedia applications are often associated with cross-domain knowledge transfer, where Unsupervised Domain Adaptation (UDA) can be used to reduce the domain shifts. Open Set Domain Adaptation (OSDA) aims to transfer knowledge from a well-labeled source domain to an unlabeled target domain under the assumption that the target domain contains unknown classes. Existing OSDA methods consistently lay stress on the covariate shift, ignoring the potential label shift problem. The performance of OSDA methods degrades drastically under intra-domain class imbalance and inter-domain label shift. However, little attention has been paid to this issue in the community. In this paper, the Imbalanced Open Set Domain Adaptation (IOSDA) is explored where the covariate shift, label shift and category mismatch exist simultaneously. To alleviate the negative effects raised by label shift in OSDA, we propose Open-set Moving-threshold Estimation and Gradual Alignment (OMEGA) - a novel architecture that improves existing OSDA methods on class-imbalanced data. Specifically, a novel unknown-aware target clustering scheme is proposed to form tight clusters in the target domain to reduce the negative effects of label shift and intra-domain class imbalance. Furthermore, moving-threshold estimation is designed to generate specific thresholds for each target sample rather than using one for all. Extensive experiments on IOSDA, OSDA and OPDA benchmarks demonstrate that our method could significantly outperform existing state-of-the-arts. Code and data are available at https://github.com/mendicant04/OMEGA.
comment: 11 pages, 5 figures, 7 tables
♻ ☆ Neighborhood Consensus Contrastive Learning for Backward-Compatible Representation AAAI 2022
In object re-identification (ReID), the development of deep learning techniques often involves model updates and deployment. It is unbearable to re-embedding and re-index with the system suspended when deploying new models. Therefore, backward-compatible representation is proposed to enable "new" features to be compared with "old" features directly, which means that the database is active when there are both "new" and "old" features in it. Thus we can scroll-refresh the database or even do nothing on the database to update. The existing backward-compatible methods either require a strong overlap between old and new training data or simply conduct constraints at the instance level. Thus they are difficult in handling complicated cluster structures and are limited in eliminating the impact of outliers in old embeddings, resulting in a risk of damaging the discriminative capability of new features. In this work, we propose a Neighborhood Consensus Contrastive Learning (NCCL) method. With no assumptions about the new training data, we estimate the sub-cluster structures of old embeddings. A new embedding is constrained with multiple old embeddings in both embedding space and discrimination space at the sub-class level. The effect of outliers diminished, as the multiple samples serve as "mean teachers". Besides, we also propose a scheme to filter the old embeddings with low credibility, further improving the compatibility robustness. Our method ensures backward compatibility without impairing the accuracy of the new model. And it can even improve the new model's accuracy in most scenarios.
comment: Accepted by AAAI 2022
♻ ☆ Pishgu: Universal Path Prediction Network Architecture for Real-time Cyber-physical Edge Systems
Path prediction is an essential task for many real-world Cyber-Physical Systems (CPS) applications, from autonomous driving and traffic monitoring/management to pedestrian/worker safety. These real-world CPS applications need a robust, lightweight path prediction that can provide a universal network architecture for multiple subjects (e.g., pedestrians and vehicles) from different perspectives. However, most existing algorithms are tailor-made for a unique subject with a specific camera perspective and scenario. This article presents Pishgu, a universal lightweight network architecture, as a robust and holistic solution for path prediction. Pishgu's architecture can adapt to multiple path prediction domains with different subjects (vehicles, pedestrians), perspectives (bird's-eye, high-angle), and scenes (sidewalk, highway). Our proposed architecture captures the inter-dependencies within the subjects in each frame by taking advantage of Graph Isomorphism Networks and the attention module. We separately train and evaluate the efficacy of our architecture on three different CPS domains across multiple perspectives (vehicle bird's-eye view, pedestrian bird's-eye view, and human high-angle view). Pishgu outperforms state-of-the-art solutions in the vehicle bird's-eye view domain by 42% and 61% and pedestrian high-angle view domain by 23% and 22% in terms of ADE and FDE, respectively. Additionally, we analyze the domain-specific details for various datasets to understand their effect on path prediction and model interpretation. Finally, we report the latency and throughput for all three domains on multiple embedded platforms showcasing the robustness and adaptability of Pishgu for real-world integration into CPS applications.
♻ ☆ Post-training Quantization on Diffusion Models CVPR 2023
Denoising diffusion (score-based) generative models have recently achieved significant accomplishments in generating realistic and diverse data. These approaches define a forward diffusion process for transforming data into noise and a backward denoising process for sampling data from noise. Unfortunately, the generation process of current denoising diffusion models is notoriously slow due to the lengthy iterative noise estimations, which rely on cumbersome neural networks. It prevents the diffusion models from being widely deployed, especially on edge devices. Previous works accelerate the generation process of diffusion model (DM) via finding shorter yet effective sampling trajectories. However, they overlook the cost of noise estimation with a heavy network in every iteration. In this work, we accelerate generation from the perspective of compressing the noise estimation network. Due to the difficulty of retraining DMs, we exclude mainstream training-aware compression paradigms and introduce post-training quantization (PTQ) into DM acceleration. However, the output distributions of noise estimation networks change with time-step, making previous PTQ methods fail in DMs since they are designed for single-time step scenarios. To devise a DM-specific PTQ method, we explore PTQ on DM in three aspects: quantized operations, calibration dataset, and calibration metric. We summarize and use several observations derived from all-inclusive investigations to formulate our method, which especially targets the unique multi-time-step structure of DMs. Experimentally, our method can directly quantize full-precision DMs into 8-bit models while maintaining or even improving their performance in a training-free manner. Importantly, our method can serve as a plug-and-play module on other fast-sampling methods, e.g., DDIM.
comment: Accepted to CVPR 2023
♻ ☆ Adaptive Activation-based Structured Pruning
Pruning is a promising approach to compress complex deep learning models in order to deploy them on resource-constrained edge devices. However, many existing pruning solutions are based on unstructured pruning, which yields models that cannot efficiently run on commodity hardware and require users to manually explore and tune the pruning process, which is time-consuming and often leads to sub-optimal results. To address these limitations, this paper presents an adaptive, activation-based, structured pruning approach to automatically and efficiently generate small, accurate, and hardware-efficient models that meet user requirements. First, it proposes iterative structured pruning using activation-based attention feature maps to effectively identify and prune unimportant filters. Then, it proposes adaptive pruning policies for automatically meeting the pruning objectives of accuracy-critical, memory-constrained, and latency-sensitive tasks. A comprehensive evaluation shows that the proposed method can substantially outperform the state-of-the-art structured pruning works on CIFAR-10 and ImageNet datasets. For example, on ResNet-56 with CIFAR-10, without any accuracy drop, our method achieves the largest parameter reduction (79.11%), outperforming the related works by 22.81% to 66.07%, and the largest FLOPs reduction (70.13%), outperforming the related works by 14.13% to 26.53%.
♻ ☆ Bringing Online Egocentric Action Recognition into the wild
To enable a safe and effective human-robot cooperation, it is crucial to develop models for the identification of human activities. Egocentric vision seems to be a viable solution to solve this problem, and therefore many works provide deep learning solutions to infer human actions from first person videos. However, although very promising, most of these do not consider the major challenges that comes with a realistic deployment, such as the portability of the model, the need for real-time inference, and the robustness with respect to the novel domains (i.e., new spaces, users, tasks). With this paper, we set the boundaries that egocentric vision models should consider for realistic applications, defining a novel setting of egocentric action recognition in the wild, which encourages researchers to develop novel, applications-aware solutions. We also present a new model-agnostic technique that enables the rapid repurposing of existing architectures in this new context, demonstrating the feasibility to deploy a model on a tiny device (Jetson Nano) and to perform the task directly on the edge with very low energy consumption (2.4W on average at 50 fps). The code is publicly available at: https://github.com/EgocentricVision/EgoWild.
comment: Accepted to RA-L, for associated video, see https://www.youtube.com/watch?v=7rtynmoYnuw&t=9s
Information Retrieval
☆ Algorithmic neutrality
Bias infects the algorithms that wield increasing control over our lives. Predictive policing systems overestimate crime in communities of color; hiring algorithms dock qualified female candidates; and facial recognition software struggles to recognize dark-skinned faces. Algorithmic bias has received significant attention. Algorithmic neutrality, in contrast, has been largely neglected. Algorithmic neutrality is my topic. I take up three questions. What is algorithmic neutrality? Is algorithmic neutrality possible? When we have an eye to algorithmic neutrality, what can we learn about algorithmic bias? To answer these questions in concrete terms, I work with a case study: search engines. Drawing on work about neutrality in science, I say that a search engine is neutral only if certain values, like political ideologies or the financial interests of the search engine operator, play no role in how the search engine ranks pages. Search neutrality, I argue, is impossible. Its impossibility seems to threaten the significance of search bias: if no search engine is neutral, then every search engine is biased. To defuse this threat, I distinguish two forms of bias, failing-on-its-own-terms bias and other-values bias. This distinction allows us to make sense of search bias, and capture its normative complexion, despite the impossibility of neutrality.
comment: 23 pages
☆ Improving Recommendation Systems with User Personality Inferred from Product Reviews WSDM'23
Personality is a psychological factor that reflects people's preferences, which in turn influences their decision-making. We hypothesize that accurate modeling of users' personalities improves recommendation systems' performance. However, acquiring such personality profiles is both sensitive and expensive. We address this problem by introducing a novel method to automatically extract personality profiles from public product review text. We then design and assess three context-aware recommendation architectures that leverage the profiles to test our hypothesis. Experiments on our two newly contributed personality datasets -- Amazon-beauty and Amazon-music -- validate our hypothesis, showing performance boosts of 3--28%.Our analysis uncovers that varying personality types contribute differently to recommendation performance: open and extroverted personalities are most helpful in music recommendation, while a conscientious personality is most helpful in beauty product recommendation. The dataset is available at https://github.com/XinyuanLu00/IRS-WSDM2023-personality-dataset.
comment: Accepted by IRS@WSDM'23
☆ Generalization analysis of an unfolding network for analysis-based Compressed Sensing
Unfolding networks have shown promising results in the Compressed Sensing (CS) field. Yet, the investigation of their generalization ability is still in its infancy. In this paper, we perform generalization analysis of a state-of-the-art ADMM-based unfolding network, which jointly learns a decoder for CS and a sparsifying redundant analysis operator. To this end, we first impose a structural constraint on the learnable sparsifier, which parametrizes the network's hypothesis class. For the latter, we estimate its Rademacher complexity. With this estimate in hand, we deliver generalization error bounds for the examined network. Finally, the validity of our theory is assessed and numerical comparisons to a state-of-the-art unfolding network are made, on synthetic and real-world datasets. Our experimental results demonstrate that our proposed framework complies with our theoretical findings and outperforms the baseline, consistently for all datasets.
☆ Evaluating the Robustness of Conversational Recommender Systems by Adversarial Examples
Conversational recommender systems (CRSs) are improving rapidly, according to the standard recommendation accuracy metrics. However, it is essential to make sure that these systems are robust in interacting with users including regular and malicious users who want to attack the system by feeding the system modified input data. In this paper, we propose an adversarial evaluation scheme including four scenarios in two categories and automatically generate adversarial examples to evaluate the robustness of these systems in the face of different input data. By executing these adversarial examples we can compare the ability of different conversational recommender systems to satisfy the user's preferences. We evaluate three CRSs by the proposed adversarial examples on two datasets. Our results show that none of these systems are robust and reliable to the adversarial examples.
comment: 10 pages
☆ Modelling Projection Bias in Intertemporal Choices: A Prospect Theory Based Approach
Users often face bundle promotions when purchasing, where they have to select between two options: buy the single item at full price, or buy the bundle at a discount. In this scenario, users' preferences are usually influenced by the projection bias, that is, users often believe that their future preferences are similar to their current preferences, causing them to make irrational and short-sighted decisions. It is of great significance to analyze the effect of the projection bias on users' preferences, and this study may help understand users' decision-making process and provide bundling and pricing strategies for sellers. Prior works typically use a linear bias model for qualitative analysis, and they cannot quantitatively calculate users' nonlinear and personalized bias. In this work, we propose Pobe, a projection bias-embedded preference model to accurately predict users' choices. The proposed Pobe introduces the prospect theory to analyze users' irrational decisions, and utilizes the weight function to handle users' nonlinear and personalized bias. Based on the proposed Pobe, we also study the impact of items' correlations or discount prices on users' choices, and provide four bundling strategies. Experimental results show that the proposed method can achieve better performance than prior works, especially when only small data is available.
♻ ☆ Complex QA and language models hybrid architectures, Survey
This paper reviews the state-of-the-art of hybrid language models architectures and strategies for "complex" question-answering (QA, CQA, CPS). Large Language Models (LLM) are good at leveraging public data on standard problems but once you want to tackle more specific complex questions or problems you may need specific architecture, knowledge, skills, methods, sensitive data protection, explainability, human approval and versatile feedback... We identify key elements augmenting LLM to solve complex questions or problems. We extend findings from the robust community edited research papers BIG, BLOOM and HELM which open source, benchmark and analyze limits and challenges of LLM in terms of tasks complexity and strict evaluation on accuracy (e.g. fairness, robustness, toxicity, ...). Recent projects like ChatGPT and GALACTICA have allowed non-specialists to grasp the great potential as well as the equally strong limitations of language models in complex QA. Hybridizing these models with different components could allow to overcome these different limits and go much further. We discuss some challenges associated with complex QA, including domain adaptation, decomposition and efficient multi-step QA, long form and non-factoid QA, safety and multi-sensitivity data protection, multimodal search, hallucinations, explainability and truthfulness, temproal reasoning. Therefore, we analyze current solutions and promising research trends, using elements such as: hybrid LLM architectures, active human reinforcement learning supervised with AI, prompting adaptation, neuro-symbolic and structured knowledge grounding, program synthesis, iterated decomposition and others.
♻ ☆ A Next Basket Recommendation Reality Check
The goal of a next basket recommendation (NBR) system is to recommend items for the next basket for a user, based on the sequence of their prior baskets. Recently, a number of methods with complex modules have been proposed that claim state-of-the-art performance. They rarely look into the predicted basket and just provide intuitive reasons for the observed improvements, e.g., better representation, capturing intentions or relations, etc. We provide a novel angle on the evaluation of next basket recommendation methods, centered on the distinction between repetition and exploration: the next basket is typically composed of previously consumed items (i.e., repeat items) and new items (i.e, explore items). We propose a set of metrics that measure the repeat/explore ratio and performance of NBR models. Using these new metrics, we analyze state-of-the-art NBR models. The results of our analysis help to clarify the extent of the actual progress achieved by existing NBR methods as well as the underlying reasons for the improvements. Overall, our work sheds light on the evaluation problem of NBR and provides useful insights into the model design for this task.
comment: This paper has been accepted to ACM TOIS
♻ ☆ A Survey on Federated Recommendation Systems
Federated learning has recently been applied to recommendation systems to protect user privacy. In federated learning settings, recommendation systems can train recommendation models only collecting the intermediate parameters instead of the real user data, which greatly enhances the user privacy. Beside, federated recommendation systems enable to collaborate with other data platforms to improve recommended model performance while meeting the regulation and privacy constraints. However, federated recommendation systems faces many new challenges such as privacy, security, heterogeneity and communication costs. While significant research has been conducted in these areas, gaps in the surveying literature still exist. In this survey, we-(1) summarize some common privacy mechanisms used in federated recommendation systems and discuss the advantages and limitations of each mechanism; (2) review some robust aggregation strategies and several novel attacks against security; (3) summarize some approaches to address heterogeneity and communication costs problems; (4)introduce some open source platforms that can be used to build federated recommendation systems; (5) present some prospective research directions in the future. This survey can guide researchers and practitioners understand the research progress in these areas.
Machine Learning
☆ PAC-NeRF: Physics Augmented Continuum Neural Radiance Fields for Geometry-Agnostic System Identification ICLR 2023
Existing approaches to system identification (estimating the physical parameters of an object) from videos assume known object geometries. This precludes their applicability in a vast majority of scenes where object geometries are complex or unknown. In this work, we aim to identify parameters characterizing a physical system from a set of multi-view videos without any assumption on object geometry or topology. To this end, we propose "Physics Augmented Continuum Neural Radiance Fields" (PAC-NeRF), to estimate both the unknown geometry and physical parameters of highly dynamic objects from multi-view videos. We design PAC-NeRF to only ever produce physically plausible states by enforcing the neural radiance field to follow the conservation laws of continuum mechanics. For this, we design a hybrid Eulerian-Lagrangian representation of the neural radiance field, i.e., we use the Eulerian grid representation for NeRF density and color fields, while advecting the neural radiance fields via Lagrangian particles. This hybrid Eulerian-Lagrangian representation seamlessly blends efficient neural rendering with the material point method (MPM) for robust differentiable physics simulation. We validate the effectiveness of our proposed framework on geometry and physical parameter estimation over a vast range of materials, including elastic bodies, plasticine, sand, Newtonian and non-Newtonian fluids, and demonstrate significant performance gain on most tasks.
comment: ICLR 2023 Spotlight. Project page: https://sites.google.com/view/PAC-NeRF
☆ Planning with Large Language Models for Code Generation ICLR 2023
Existing large language model-based code generation pipelines typically use beam search or sampling algorithms during the decoding process. Although the programs they generate achieve high token-matching-based scores, they often fail to compile or generate incorrect outputs. The main reason is that conventional Transformer decoding algorithms may not be the best choice for code generation. In this work, we propose a novel Transformer decoding algorithm, Planning-Guided Transformer Decoding (PG-TD), that uses a planning algorithm to do lookahead search and guide the Transformer to generate better programs. Specifically, instead of simply optimizing the likelihood of the generated sequences, the Transformer makes use of a planner to generate candidate programs and test them on public test cases. The Transformer can therefore make more informed decisions and generate tokens that will eventually lead to higher-quality programs. We also design a mechanism that shares information between the Transformer and the planner to make our algorithm computationally efficient. We empirically evaluate our framework with several large language models as backbones on public coding challenge benchmarks, showing that 1) it can generate programs that consistently achieve higher performance compared with competing baseline methods; 2) it enables controllable code generation, such as concise codes and highly-commented codes by optimizing modified objective.
comment: ICLR 2023. Project page:https://codeaimcts.github.io
☆ Scaling up GANs for Text-to-Image Synthesis CVPR 2023
The recent success of text-to-image synthesis has taken the world by storm and captured the general public's imagination. From a technical standpoint, it also marked a drastic change in the favored architecture to design generative image models. GANs used to be the de facto choice, with techniques like StyleGAN. With DALL-E 2, auto-regressive and diffusion models became the new standard for large-scale generative models overnight. This rapid shift raises a fundamental question: can we scale up GANs to benefit from large datasets like LAION? We find that na\"Ively increasing the capacity of the StyleGAN architecture quickly becomes unstable. We introduce GigaGAN, a new GAN architecture that far exceeds this limit, demonstrating GANs as a viable option for text-to-image synthesis. GigaGAN offers three major advantages. First, it is orders of magnitude faster at inference time, taking only 0.13 seconds to synthesize a 512px image. Second, it can synthesize high-resolution images, for example, 16-megapixel pixels in 3.66 seconds. Finally, GigaGAN supports various latent space editing applications such as latent interpolation, style mixing, and vector arithmetic operations.
comment: CVPR 2023. Project webpage at https://mingukkang.github.io/GigaGAN/
☆ TANGOS: Regularizing Tabular Neural Networks through Gradient Orthogonalization and Specialization ICLR
Despite their success with unstructured data, deep neural networks are not yet a panacea for structured tabular data. In the tabular domain, their efficiency crucially relies on various forms of regularization to prevent overfitting and provide strong generalization performance. Existing regularization techniques include broad modelling decisions such as choice of architecture, loss functions, and optimization methods. In this work, we introduce Tabular Neural Gradient Orthogonalization and Specialization (TANGOS), a novel framework for regularization in the tabular setting built on latent unit attributions. The gradient attribution of an activation with respect to a given input feature suggests how the neuron attends to that feature, and is often employed to interpret the predictions of deep networks. In TANGOS, we take a different approach and incorporate neuron attributions directly into training to encourage orthogonalization and specialization of latent attributions in a fully-connected network. Our regularizer encourages neurons to focus on sparse, non-overlapping input features and results in a set of diverse and specialized latent units. In the tabular domain, we demonstrate that our approach can lead to improved out-of-sample generalization performance, outperforming other popular regularization methods. We provide insight into why our regularizer is effective and demonstrate that TANGOS can be applied jointly with existing methods to achieve even greater generalization performance.
comment: Published at International Conference on Learning Representations (ICLR) 2023
☆ Computable Phenotypes to Characterize Changing Patient Brain Dysfunction in the Intensive Care Unit
In the United States, more than 5 million patients are admitted annually to ICUs, with ICU mortality of 10%-29% and costs over $82 billion. Acute brain dysfunction status, delirium, is often underdiagnosed or undervalued. This study's objective was to develop automated computable phenotypes for acute brain dysfunction states and describe transitions among brain dysfunction states to illustrate the clinical trajectories of ICU patients. We created two single-center, longitudinal EHR datasets for 48,817 adult patients admitted to an ICU at UFH Gainesville (GNV) and Jacksonville (JAX). We developed algorithms to quantify acute brain dysfunction status including coma, delirium, normal, or death at 12-hour intervals of each ICU admission and to identify acute brain dysfunction phenotypes using continuous acute brain dysfunction status and k-means clustering approach. There were 49,770 admissions for 37,835 patients in UFH GNV dataset and 18,472 admissions for 10,982 patients in UFH JAX dataset. In total, 18% of patients had coma as the worst brain dysfunction status; every 12 hours, around 4%-7% would transit to delirium, 22%-25% would recover, 3%-4% would expire, and 67%-68% would remain in a coma in the ICU. Additionally, 7% of patients had delirium as the worst brain dysfunction status; around 6%-7% would transit to coma, 40%-42% would be no delirium, 1% would expire, and 51%-52% would remain delirium in the ICU. There were three phenotypes: persistent coma/delirium, persistently normal, and transition from coma/delirium to normal almost exclusively in first 48 hours after ICU admission. We developed phenotyping scoring algorithms that determined acute brain dysfunction status every 12 hours while admitted to the ICU. This approach may be useful in developing prognostic and decision-support tools to aid patients and clinicians in decision-making on resource use and escalation of care.
comment: 21 pages, 5 figures, 3 tables, 1 eTable
☆ Open-world Instance Segmentation: Top-down Learning with Bottom-up Supervision
Many top-down architectures for instance segmentation achieve significant success when trained and tested on pre-defined closed-world taxonomy. However, when deployed in the open world, they exhibit notable bias towards seen classes and suffer from significant performance drop. In this work, we propose a novel approach for open world instance segmentation called bottom-Up and top-Down Open-world Segmentation (UDOS) that combines classical bottom-up segmentation algorithms within a top-down learning framework. UDOS first predicts parts of objects using a top-down network trained with weak supervision from bottom-up segmentations. The bottom-up segmentations are class-agnostic and do not overfit to specific taxonomies. The part-masks are then fed into affinity-based grouping and refinement modules to predict robust instance-level segmentations. UDOS enjoys both the speed and efficiency from the top-down architectures and the generalization ability to unseen categories from bottom-up supervision. We validate the strengths of UDOS on multiple cross-category as well as cross-dataset transfer tasks from 5 challenging datasets including MS-COCO, LVIS, ADE20k, UVO and OpenImages, achieving significant improvements over state-of-the-art across the board. Our code and models are available on our project page.
comment: Project page: https://tarun005.github.io/UDOS
☆ PDSketch: Integrated Planning Domain Programming and Learning NeurIPS 2022
This paper studies a model learning and online planning approach towards building flexible and general robots. Specifically, we investigate how to exploit the locality and sparsity structures in the underlying environmental transition model to improve model generalization, data-efficiency, and runtime-efficiency. We present a new domain definition language, named PDSketch. It allows users to flexibly define high-level structures in the transition models, such as object and feature dependencies, in a way similar to how programmers use TensorFlow or PyTorch to specify kernel sizes and hidden dimensions of a convolutional neural network. The details of the transition model will be filled in by trainable neural networks. Based on the defined structures and learned parameters, PDSketch automatically generates domain-independent planning heuristics without additional training. The derived heuristics accelerate the performance-time planning for novel goals.
comment: NeurIPS 2022. Project page: https://pdsketch.csail.mit.edu
☆ Mark My Words: Dangers of Watermarked Images in ImageNet ICLR 2023
The utilization of pre-trained networks, especially those trained on ImageNet, has become a common practice in Computer Vision. However, prior research has indicated that a significant number of images in the ImageNet dataset contain watermarks, making pre-trained networks susceptible to learning artifacts such as watermark patterns within their latent spaces. In this paper, we aim to assess the extent to which popular pre-trained architectures display such behavior and to determine which classes are most affected. Additionally, we examine the impact of watermarks on the extracted features. Contrary to the popular belief that the Chinese logographic watermarks impact the "carton" class only, our analysis reveals that a variety of ImageNet classes, such as "monitor", "broom", "apron" and "safe" rely on spurious correlations. Finally, we propose a simple approach to mitigate this issue in fine-tuned networks by ignoring the encodings from the feature-extractor layer of ImageNet pre-trained networks that are most susceptible to watermark imprints.
comment: 5 pages, 4 figures, Accepted to the ICLR 2023 TrustML-(un)Limited workshop
☆ Learning Stationary Markov Processes with Contrastive Adjustment
We introduce a new optimization algorithm, termed \emph{contrastive adjustment}, for learning Markov transition kernels whose stationary distribution matches the data distribution. Contrastive adjustment is not restricted to a particular family of transition distributions and can be used to model data in both continuous and discrete state spaces. Inspired by recent work on noise-annealed sampling, we propose a particular transition operator, the \emph{noise kernel}, that can trade mixing speed for sample fidelity. We show that contrastive adjustment is highly valuable in human-computer design processes, as the stationarity of the learned Markov chain enables local exploration of the data manifold and makes it possible to iteratively refine outputs by human feedback. We compare the performance of noise kernels trained with contrastive adjustment to current state-of-the-art generative models and demonstrate promising results on a variety of image synthesis tasks.
☆ Sparse and Local Networks for Hypergraph Reasoning
Reasoning about the relationships between entities from input facts (e.g., whether Ari is a grandparent of Charlie) generally requires explicit consideration of other entities that are not mentioned in the query (e.g., the parents of Charlie). In this paper, we present an approach for learning to solve problems of this kind in large, real-world domains, using sparse and local hypergraph neural networks (SpaLoc). SpaLoc is motivated by two observations from traditional logic-based reasoning: relational inferences usually apply locally (i.e., involve only a small number of individuals), and relations are usually sparse (i.e., only hold for a small percentage of tuples in a domain). We exploit these properties to make learning and inference efficient in very large domains by (1) using a sparse tensor representation for hypergraph neural networks, (2) applying a sparsification loss during training to encourage sparse representations, and (3) subsampling based on a novel information sufficiency-based sampling process during training. SpaLoc achieves state-of-the-art performance on several real-world, large-scale knowledge graph reasoning benchmarks, and is the first framework for applying hypergraph neural networks on real-world knowledge graphs with more than 10k nodes.
comment: Learning on Graphs Conference (LoG) 2022. Project page: https://spaloc.csail.mit.edu
☆ On the Expressiveness and Generalization of Hypergraph Neural Networks
This extended abstract describes a framework for analyzing the expressiveness, learning, and (structural) generalization of hypergraph neural networks (HyperGNNs). Specifically, we focus on how HyperGNNs can learn from finite datasets and generalize structurally to graph reasoning problems of arbitrary input sizes. Our first contribution is a fine-grained analysis of the expressiveness of HyperGNNs, that is, the set of functions that they can realize. Our result is a hierarchy of problems they can solve, defined in terms of various hyperparameters such as depths and edge arities. Next, we analyze the learning properties of these neural networks, especially focusing on how they can be trained on a finite set of small graphs and generalize to larger graphs, which we term structural generalization. Our theoretical results are further supported by the empirical results.
comment: Learning on Graphs Conference (LoG) 2022
☆ Learning Rational Subgoals from Demonstrations and Instructions AAAI 2023
We present a framework for learning useful subgoals that support efficient long-term planning to achieve novel goals. At the core of our framework is a collection of rational subgoals (RSGs), which are essentially binary classifiers over the environmental states. RSGs can be learned from weakly-annotated data, in the form of unsegmented demonstration trajectories, paired with abstract task descriptions, which are composed of terms initially unknown to the agent (e.g., collect-wood then craft-boat then go-across-river). Our framework also discovers dependencies between RSGs, e.g., the task collect-wood is a helpful subgoal for the task craft-boat. Given a goal description, the learned subgoals and the derived dependencies facilitate off-the-shelf planning algorithms, such as A* and RRT, by setting helpful subgoals as waypoints to the planner, which significantly improves performance-time efficiency.
comment: AAAI 2023. First two authors contributed equally. Project page: https://rsg.csail.mit.edu
☆ Efficient Testable Learning of Halfspaces with Adversarial Label Noise
We give the first polynomial-time algorithm for the testable learning of halfspaces in the presence of adversarial label noise under the Gaussian distribution. In the recently introduced testable learning model, one is required to produce a tester-learner such that if the data passes the tester, then one can trust the output of the robust learner on the data. Our tester-learner runs in time $\poly(d/\eps)$ and outputs a halfspace with misclassification error $O(\opt)+\eps$, where $\opt$ is the 0-1 error of the best fitting halfspace. At a technical level, our algorithm employs an iterative soft localization technique enhanced with appropriate testers to ensure that the data distribution is sufficiently similar to a Gaussian.
☆ Cal-QL: Calibrated Offline RL Pre-Training for Efficient Online Fine-Tuning
A compelling use case of offline reinforcement learning (RL) is to obtain a policy initialization from existing datasets, which allows efficient fine-tuning with limited amounts of active online interaction. However, several existing offline RL methods tend to exhibit poor online fine-tuning performance. On the other hand, online RL methods can learn effectively through online interaction, but struggle to incorporate offline data, which can make them very slow in settings where exploration is challenging or pre-training is necessary. In this paper, we devise an approach for learning an effective initialization from offline data that also enables fast online fine-tuning capabilities. Our approach, calibrated Q-learning (Cal-QL) accomplishes this by learning a conservative value function initialization that underestimates the value of the learned policy from offline data, while also being calibrated, in the sense that the learned Q-values are at a reasonable scale. We refer to this property as calibration, and define it formally as providing a lower bound on the true value function of the learned policy and an upper bound on the value of some other (suboptimal) reference policy, which may simply be the behavior policy. We show that offline RL algorithms that learn such calibrated value functions lead to effective online fine-tuning, enabling us to take the benefits of offline initializations in online fine-tuning. In practice, Cal-QL can be implemented on top of existing conservative methods for offline RL within a one-line code change. Empirically, Cal-QL outperforms state-of-the-art methods on 10/11 fine-tuning benchmark tasks that we study in this paper.
☆ Spawrious: A Benchmark for Fine Control of Spurious Correlation Biases
The problem of spurious correlations (SCs) arises when a classifier relies on non-predictive features that happen to be correlated with the labels in the training data. For example, a classifier may misclassify dog breeds based on the background of dog images. This happens when the backgrounds are correlated with other breeds in the training data, leading to misclassifications during test time. Previous SC benchmark datasets suffer from varying issues, e.g., over-saturation or only containing one-to-one (O2O) SCs, but no many-to-many (M2M) SCs arising between groups of spurious attributes and classes. In this paper, we present Spawrious-{O2O, M2M}-{Easy, Medium, Hard}, an image classification benchmark suite containing spurious correlations among different dog breeds and background locations. To create this dataset, we employ a text-to-image model to generate photo-realistic images, and an image captioning model to filter out unsuitable ones. The resulting dataset is of high quality, containing approximately 152,000 images. Our experimental results demonstrate that state-of-the-art group robustness methods struggle with Spawrious, most notably on the Hard-splits with $<60\%$ accuracy. By examining model misclassifications, we detect reliances on spurious backgrounds, demonstrating that our dataset provides a significant challenge to drive future research.
☆ Resolving quantitative MRI model degeneracy with machine learning via training data distribution design
Quantitative MRI (qMRI) aims to map tissue properties non-invasively via models that relate these unknown quantities to measured MRI signals. Estimating these unknowns, which has traditionally required model fitting - an often iterative procedure, can now be done with one-shot machine learning (ML) approaches. Such parameter estimation may be complicated by intrinsic qMRI signal model degeneracy: different combinations of tissue properties produce the same signal. Despite their many advantages, it remains unclear whether ML approaches can resolve this issue. Growing empirical evidence appears to suggest ML approaches remain susceptible to model degeneracy. Here we demonstrate under the right circumstances ML can address this issue. Inspired by recent works on the impact of training data distributions on ML-based parameter estimation, we propose to resolve model degeneracy by designing training data distributions. We put forward a classification of model degeneracies and identify one particular kind of degeneracies amenable to the proposed attack. The strategy is demonstrated successfully using the Revised NODDI model with standard multi-shell diffusion MRI data as an exemplar. Our results illustrate the importance of training set design which has the potential to allow accurate estimation of tissue properties with ML.
comment: 13 pages, 4 figures
☆ Beware of Instantaneous Dependence in Reinforcement Learning
Playing an important role in Model-Based Reinforcement Learning (MBRL), environment models aim to predict future states based on the past. Existing works usually ignore instantaneous dependence in the state, that is, assuming that the future state variables are conditionally independent given the past states. However, instantaneous dependence is prevalent in many RL environments. For instance, in the stock market, instantaneous dependence can exist between two stocks because the fluctuation of one stock can quickly affect the other and the resolution of price change is lower than that of the effect. In this paper, we prove that with few exceptions, ignoring instantaneous dependence can result in suboptimal policy learning in MBRL. To address the suboptimality problem, we propose a simple plug-and-play method to enable existing MBRL algorithms to take instantaneous dependence into account. Through experiments on two benchmarks, we (1) confirm the existence of instantaneous dependence with visualization; (2) validate our theoretical findings that ignoring instantaneous dependence leads to suboptimal policy; (3) verify that our method effectively enables reinforcement learning with instantaneous dependence and improves policy performance.
☆ Power and Interference Control for VLC-Based UDN: A Reinforcement Learning Approach ISWC'2022
Visible light communication (VLC) has been widely applied as a promising solution for modern short range communication. When it comes to the deployment of LED arrays in VLC networks, the emerging ultra-dense network (UDN) technology can be adopted to expand the VLC network's capacity. However, the problem of inter-cell interference (ICI) mitigation and efficient power control in the VLC-based UDN is still a critical challenge. To this end, a reinforcement learning (RL) based VLC UDN architecture is devised in this paper. The deployment of the cells is optimized via spatial reuse to mitigate ICI. An RL-based algorithm is proposed to dynamically optimize the policy of power and interference control, maximizing the system utility in the complicated and dynamic environment. Simulation results demonstrate the superiority of the proposed scheme, it increase the system utility and achievable data rate while reducing the energy consumption and ICI, which outperforms the benchmark scheme.
comment: This paper has been accepted by and to appear in Proc. ACM UbiComp/ISWC'2022
☆ Communication-Efficient Collaborative Heterogeneous Bandits in Networks
The multi-agent multi-armed bandit problem has been studied extensively due to its ubiquity in many real-life applications, such as online recommendation systems and wireless networking. We consider the setting where agents should minimize their group regret while collaborating over a given graph via some communication protocol and where each agent is given a different set of arms. Previous literature on this problem only considered one of the two desired features separately: agents with the same arm set communicate over a general graph, or agents with different arm sets communicate over a fully connected graph. In this work, we introduce a more general problem setting that encompasses all the desired features. For this novel setting, we first provide a rigorous regret analysis for the standard flooding protocol combined with the UCB policy. Then, to mitigate the issue of high communication costs incurred by flooding, we propose a new protocol called Flooding with Absorption (FWA). We provide a theoretical analysis of the regret bound and intuitions on the advantages of using FWA over flooding. Lastly, we verify empirically that using FWA leads to significantly lower communication costs despite minimal regret performance loss compared to flooding.
comment: 17 pages, 7 figures; submitted to MobiHoc 2023
☆ Quantum Splines for Non-Linear Approximations
Quantum Computing offers a new paradigm for efficient computing and many AI applications could benefit from its potential boost in performance. However, the main limitation is the constraint to linear operations that hampers the representation of complex relationships in data. In this work, we propose an efficient implementation of quantum splines for non-linear approximation. In particular, we first discuss possible parametrisations, and select the most convenient for exploiting the HHL algorithm to obtain the estimates of spline coefficients. Then, we investigate QSpline performance as an evaluation routine for some of the most popular activation functions adopted in ML. Finally, a detailed comparison with classical alternatives to the HHL is also presented.
comment: 6 pages, 3 figures, 2 tables
☆ Kernel Regression with Infinite-Width Neural Networks on Millions of Examples
Neural kernels have drastically increased performance on diverse and nonstandard data modalities but require significantly more compute, which previously limited their application to smaller datasets. In this work, we address this by massively parallelizing their computation across many GPUs. We combine this with a distributed, preconditioned conjugate gradients algorithm to enable kernel regression at a large scale (i.e. up to five million examples). Using this approach, we study scaling laws of several neural kernels across many orders of magnitude for the CIFAR-5m dataset. Using data augmentation to expand the original CIFAR-10 training dataset by a factor of 20, we obtain a test accuracy of 91.2\% (SotA for a pure kernel method). Moreover, we explore neural kernels on other data modalities, obtaining results on protein and small molecule prediction tasks that are competitive with SotA methods.
☆ Fast kernel methods for Data Quality Monitoring as a goodness-of-fit test
We here propose a machine learning approach for monitoring particle detectors in real-time. The goal is to assess the compatibility of incoming experimental data with a reference dataset, characterising the data behaviour under normal circumstances, via a likelihood-ratio hypothesis test. The model is based on a modern implementation of kernel methods, nonparametric algorithms that can learn any continuous function given enough data. The resulting approach is efficient and agnostic to the type of anomaly that may be present in the data. Our study demonstrates the effectiveness of this strategy on multivariate data from drift tube chamber muon detectors.
comment: 16 pages, 7 figures
☆ Greener yet Powerful: Taming Large Code Generation Models with Quantization
ML-powered code generation aims to assist developers to write code in a more productive manner, by intelligently generating code blocks based on natural language prompts. Recently, large pretrained deep learning models have substantially pushed the boundary of code generation and achieved impressive performance. Despite their great power, the huge number of model parameters poses a significant threat to adapting them in a regular software development environment, where a developer might use a standard laptop or mid-size server to develop her code. Such large models incur significant resource usage (in terms of memory, latency, and dollars) as well as carbon footprint. Model compression is a promising approach to address these challenges. Several techniques are proposed to compress large pretrained models typically used for vision or textual data. Out of many available compression techniques, we identified that quantization is mostly applicable for code generation task as it does not require significant retraining cost. As quantization represents model parameters with lower-bit integer (e.g., int8), the model size and runtime latency would both benefit from such int representation. We extensively study the impact of quantized model on code generation tasks across different dimension: (i) resource usage and carbon footprint, (ii) accuracy, and (iii) robustness. To this end, through systematic experiments we find a recipe of quantization technique that could run even a $6$B model in a regular laptop without significant accuracy or robustness degradation. We further found the recipe is readily applicable to code summarization task as well.
comment: 10 pages, 7 figures, 10 tables
☆ PC-JeDi: Diffusion for Particle Cloud Generation in High Energy Physics
In this paper, we present a new method to efficiently generate jets in High Energy Physics called PC-JeDi. This method utilises score-based diffusion models in conjunction with transformers which are well suited to the task of generating jets as particle clouds due to their permutation equivariance. PC-JeDi achieves competitive performance with current state-of-the-art methods across several metrics that evaluate the quality of the generated jets. Although slower than other models, due to the large number of forward passes required by diffusion models, it is still substantially faster than traditional detailed simulation. Furthermore, PC-JeDi uses conditional generation to produce jets with a desired mass and transverse momentum for two different particles, top quarks and gluons.
comment: 29 pages, 25 figures, 5 tables
☆ Data-dependent Generalization Bounds via Variable-Size Compressibility
In this paper, we establish novel data-dependent upper bounds on the generalization error through the lens of a "variable-size compressibility" framework that we introduce newly here. In this framework, the generalization error of an algorithm is linked to a variable-size 'compression rate' of its input data. This is shown to yield bounds that depend on the empirical measure of the given input data at hand, rather than its unknown distribution. Our new generalization bounds that we establish are tail bounds, tail bounds on the expectation, and in-expectations bounds. Moreover, it is shown that our framework also allows to derive general bounds on any function of the input data and output hypothesis random variables. In particular, these general bounds are shown to subsume and possibly improve over several existing PAC-Bayes and data-dependent intrinsic dimension-based bounds that are recovered as special cases, thus unveiling a unifying character of our approach. For instance, a new data-dependent intrinsic dimension based bounds is established, which connects the generalization error to the optimization trajectories and reveals various interesting connections with rate-distortion dimension of process, R\'enyi information dimension of process, and metric mean dimension.
comment: 35 pages
☆ Penalized Deep Partially Linear Cox Models with Application to CT Scans of Lung Cancer Patients
Lung cancer is a leading cause of cancer mortality globally, highlighting the importance of understanding its mortality risks to design effective patient-centered therapies. The National Lung Screening Trial (NLST) was a nationwide study aimed at investigating risk factors for lung cancer. The study employed computed tomography texture analysis (CTTA), which provides objective measurements of texture patterns on CT scans, to quantify the mortality risks of lung cancer patients. Partially linear Cox models are becoming a popular tool for modeling survival outcomes, as they effectively handle both established risk factors (such as age and other clinical factors) and new risk factors (such as image features) in a single framework. The challenge in identifying the texture features that impact cancer survival is due to their sensitivity to factors such as scanner type, segmentation, and organ motion. To overcome this challenge, we propose a novel Penalized Deep Partially Linear Cox Model (Penalized DPLC), which incorporates the SCAD penalty to select significant texture features and employs a deep neural network to estimate the nonparametric component of the model accurately. We prove the convergence and asymptotic properties of the estimator and compare it to other methods through extensive simulation studies, evaluating its performance in risk prediction and feature selection. The proposed method is applied to the NLST study dataset to uncover the effects of key clinical and imaging risk factors on patients' survival. Our findings provide valuable insights into the relationship between these factors and survival outcomes.
☆ Adaptive Calibrator Ensemble for Model Calibration under Distribution Shift
Model calibration usually requires optimizing some parameters (e.g., temperature) w.r.t an objective function (e.g., negative log-likelihood). In this paper, we report a plain, important but often neglected fact that the objective function is influenced by calibration set difficulty, i.e., the ratio of the number of incorrectly classified samples to that of correctly classified samples. If a test set has a drastically different difficulty level from the calibration set, the optimal calibration parameters of the two datasets would be different. In other words, a calibrator optimal on the calibration set would be suboptimal on the OOD test set and thus has degraded performance. With this knowledge, we propose a simple and effective method named adaptive calibrator ensemble (ACE) to calibrate OOD datasets whose difficulty is usually higher than the calibration set. Specifically, two calibration functions are trained, one for in-distribution data (low difficulty), and the other for severely OOD data (high difficulty). To achieve desirable calibration on a new OOD dataset, ACE uses an adaptive weighting method that strikes a balance between the two extreme functions. When plugged in, ACE generally improves the performance of a few state-of-the-art calibration schemes on a series of OOD benchmarks. Importantly, such improvement does not come at the cost of the in-distribution calibration accuracy.
comment: 16 pages, 9 figures
☆ CoolPINNs: A Physics-informed Neural Network Modeling of Active Cooling in Vascular Systems
Emerging technologies like hypersonic aircraft, space exploration vehicles, and batteries avail fluid circulation in embedded microvasculatures for efficient thermal regulation. Modeling is vital during these engineered systems' design and operational phases. However, many challenges exist in developing a modeling framework. What is lacking is an accurate framework that (i) captures sharp jumps in the thermal flux across complex vasculature layouts, (ii) deals with oblique derivatives (involving tangential and normal components), (iii) handles nonlinearity because of radiative heat transfer, (iv) provides a high-speed forecast for real-time monitoring, and (v) facilitates robust inverse modeling. This paper addresses these challenges by availing the power of physics-informed neural networks (PINNs). We develop a fast, reliable, and accurate Scientific Machine Learning (SciML) framework for vascular-based thermal regulation -- called CoolPINNs: a PINNs-based modeling framework for active cooling. The proposed mesh-less framework elegantly overcomes all the mentioned challenges. The significance of the reported research is multi-fold. First, the framework is valuable for real-time monitoring of thermal regulatory systems because of rapid forecasting. Second, researchers can address complex thermoregulation designs inasmuch as the approach is mesh-less. Finally, the framework facilitates systematic parameter identification and inverse modeling studies, perhaps the current framework's most significant utility.
☆ Dynamic Stashing Quantization for Efficient Transformer Training
Large Language Models (LLMs) have demonstrated impressive performance on a range of Natural Language Processing (NLP) tasks. Unfortunately, the immense amount of computations and memory accesses required for LLM training makes them prohibitively expensive in terms of hardware cost, and thus challenging to deploy in use cases such as on-device learning. In this paper, motivated by the observation that LLM training is memory-bound, we propose a novel dynamic quantization strategy, termed Dynamic Stashing Quantization (DSQ), that puts a special focus on reducing the memory operations, but also enjoys the other benefits of low precision training, such as the reduced arithmetic cost. We conduct a thorough study on two translation tasks (trained-from-scratch) and three classification tasks (fine-tuning). DSQ reduces the amount of arithmetic operations by $20.95\times$ and the number of DRAM operations by $2.55\times$ on IWSLT17 compared to the standard 16-bit fixed-point, which is widely used in on-device learning.
☆ Fast post-process Bayesian inference with Sparse Variational Bayesian Monte Carlo
We introduce Sparse Variational Bayesian Monte Carlo (SVBMC), a method for fast "post-process" Bayesian inference for models with black-box and potentially noisy likelihoods. SVBMC reuses all existing target density evaluations -- for example, from previous optimizations or partial Markov Chain Monte Carlo runs -- to build a sparse Gaussian process (GP) surrogate model of the log posterior density. Uncertain regions of the surrogate are then refined via active learning as needed. Our work builds on the Variational Bayesian Monte Carlo (VBMC) framework for sample-efficient inference, with several novel contributions. First, we make VBMC scalable to a large number of pre-existing evaluations via sparse GP regression, deriving novel Bayesian quadrature formulae and acquisition functions for active learning with sparse GPs. Second, we introduce noise shaping, a general technique to induce the sparse GP approximation to focus on high posterior density regions. Third, we prove theoretical results in support of the SVBMC refinement procedure. We validate our method on a variety of challenging synthetic scenarios and real-world applications. We find that SVBMC consistently builds good posterior approximations by post-processing of existing model evaluations from different sources, often requiring only a small number of additional density evaluations.
comment: 41 pages, 17 figures
☆ Efficient Certified Training and Robustness Verification of Neural ODEs ICLR23
Neural Ordinary Differential Equations (NODEs) are a novel neural architecture, built around initial value problems with learned dynamics which are solved during inference. Thought to be inherently more robust against adversarial perturbations, they were recently shown to be vulnerable to strong adversarial attacks, highlighting the need for formal guarantees. However, despite significant progress in robustness verification for standard feed-forward architectures, the verification of high dimensional NODEs remains an open problem. In this work, we address this challenge and propose GAINS, an analysis framework for NODEs combining three key ideas: (i) a novel class of ODE solvers, based on variable but discrete time steps, (ii) an efficient graph representation of solver trajectories, and (iii) a novel abstraction algorithm operating on this graph representation. Together, these advances enable the efficient analysis and certified training of high-dimensional NODEs, by reducing the runtime from an intractable $O(\exp(d)+\exp(T))$ to ${O}(d+T^2 \log^2T)$ in the dimensionality $d$ and integration time $T$. In an extensive evaluation on computer vision (MNIST and FMNIST) and time-series forecasting (PHYSIO-NET) problems, we demonstrate the effectiveness of both our certified training and verification methods.
comment: Accepted at ICLR23
☆ Structure-Aware Group Discrimination with Adaptive-View Graph Encoder: A Fast Graph Contrastive Learning Framework
Albeit having gained significant progress lately, large-scale graph representation learning remains expensive to train and deploy for two main reasons: (i) the repetitive computation of multi-hop message passing and non-linearity in graph neural networks (GNNs); (ii) the computational cost of complex pairwise contrastive learning loss. Two main contributions are made in this paper targeting this twofold challenge: we first propose an adaptive-view graph neural encoder (AVGE) with a limited number of message passing to accelerate the forward pass computation, and then we propose a structure-aware group discrimination (SAGD) loss in our framework which avoids inefficient pairwise loss computing in most common GCL and improves the performance of the simple group discrimination. By the framework proposed, we manage to bring down the training and inference cost on various large-scale datasets by a significant margin (250x faster inference time) without loss of the downstream-task performance.
☆ Taming Contrast Maximization for Learning Sequential, Low-latency, Event-based Optical Flow
Event cameras have recently gained significant traction since they open up new avenues for low-latency and low-power solutions to complex computer vision problems. To unlock these solutions, it is necessary to develop algorithms that can leverage the unique nature of event data. However, the current state-of-the-art is still highly influenced by the frame-based literature, and usually fails to deliver on these promises. In this work, we take this into consideration and propose a novel self-supervised learning pipeline for the sequential estimation of event-based optical flow that allows for the scaling of the models to high inference frequencies. At its core, we have a continuously-running stateful neural model that is trained using a novel formulation of contrast maximization that makes it robust to nonlinearities and varying statistics in the input events. Results across multiple datasets confirm the effectiveness of our method, which establishes a new state of the art in terms of accuracy for approaches trained or optimized without ground truth.
comment: 15 pages, 12 figures, 7 tables
☆ FedREP: A Byzantine-Robust, Communication-Efficient and Privacy-Preserving Framework for Federated Learning
Federated learning (FL) has recently become a hot research topic, in which Byzantine robustness, communication efficiency and privacy preservation are three important aspects. However, the tension among these three aspects makes it hard to simultaneously take all of them into account. In view of this challenge, we theoretically analyze the conditions that a communication compression method should satisfy to be compatible with existing Byzantine-robust methods and privacy-preserving methods. Motivated by the analysis results, we propose a novel communication compression method called consensus sparsification (ConSpar). To the best of our knowledge, ConSpar is the first communication compression method that is designed to be compatible with both Byzantine-robust methods and privacy-preserving methods. Based on ConSpar, we further propose a novel FL framework called FedREP, which is Byzantine-robust, communication-efficient and privacy-preserving. We theoretically prove the Byzantine robustness and the convergence of FedREP. Empirical results show that FedREP can significantly outperform communication-efficient privacy-preserving baselines. Furthermore, compared with Byzantine-robust communication-efficient baselines, FedREP can achieve comparable accuracy with the extra advantage of privacy preservation.
☆ Real-time scheduling of renewable power systems through planning-based reinforcement learning
The growing renewable energy sources have posed significant challenges to traditional power scheduling. It is difficult for operators to obtain accurate day-ahead forecasts of renewable generation, thereby requiring the future scheduling system to make real-time scheduling decisions aligning with ultra-short-term forecasts. Restricted by the computation speed, traditional optimization-based methods can not solve this problem. Recent developments in reinforcement learning (RL) have demonstrated the potential to solve this challenge. However, the existing RL methods are inadequate in terms of constraint complexity, algorithm performance, and environment fidelity. We are the first to propose a systematic solution based on the state-of-the-art reinforcement learning algorithm and the real power grid environment. The proposed approach enables planning and finer time resolution adjustments of power generators, including unit commitment and economic dispatch, thus increasing the grid's ability to admit more renewable energy. The well-trained scheduling agent significantly reduces renewable curtailment and load shedding, which are issues arising from traditional scheduling's reliance on inaccurate day-ahead forecasts. High-frequency control decisions exploit the existing units' flexibility, reducing the power grid's dependence on hardware transformations and saving investment and operating costs, as demonstrated in experimental results. This research exhibits the potential of reinforcement learning in promoting low-carbon and intelligent power systems and represents a solid step toward sustainable electricity generation.
comment: 12 pages, 7 figures
☆ Mastering Strategy Card Game (Hearthstone) with Improved Techniques
Strategy card game is a well-known genre that is demanding on the intelligent game-play and can be an ideal test-bench for AI. Previous work combines an end-to-end policy function and an optimistic smooth fictitious play, which shows promising performances on the strategy card game Legend of Code and Magic. In this work, we apply such algorithms to Hearthstone, a famous commercial game that is more complicated in game rules and mechanisms. We further propose several improved techniques and consequently achieve significant progress. For a machine-vs-human test we invite a Hearthstone streamer whose best rank was top 10 of the official league in China region that is estimated to be of millions of players. Our models defeat the human player in all Best-of-5 tournaments of full games (including both deck building and battle), showing a strong capability of decision making.
☆ GOATS: Goal Sampling Adaptation for Scooping with Curriculum Reinforcement Learning
In this work, we first formulate the problem of goal-conditioned robotic water scooping with reinforcement learning. This task is challenging due to the complex dynamics of fluid and multi-modal goal-reaching. The policy is required to achieve both position goals and water amount goals, which leads to a large convoluted goal state space. To address these challenges, we introduce Goal Sampling Adaptation for Scooping (GOATS), a curriculum reinforcement learning method that can learn an effective and generalizable policy for robot scooping tasks. Specifically, we use a goal-factorized reward formulation and interpolate position goal distributions and amount goal distributions to create curriculum through the learning process. As a result, our proposed method can outperform the baselines in simulation and achieves 5.46% and 8.71% amount errors on bowl scooping and bucket scooping tasks, respectively, under 1000 variations of initial water states in the tank and a large goal state space. Besides being effective in simulation environments, our method can efficiently generalize to noisy real-robot water-scooping scenarios with different physical configurations and unseen settings, demonstrating superior efficacy and generalizability. The videos of this work are available on our project page: https://sites.google.com/view/goatscooping.
☆ A Framework for History-Aware Hyperparameter Optimisation in Reinforcement Learning
A Reinforcement Learning (RL) system depends on a set of initial conditions (hyperparameters) that affect the system's performance. However, defining a good choice of hyperparameters is a challenging problem. Hyperparameter tuning often requires manual or automated searches to find optimal values. Nonetheless, a noticeable limitation is the high cost of algorithm evaluation for complex models, making the tuning process computationally expensive and time-consuming. In this paper, we propose a framework based on integrating complex event processing and temporal models, to alleviate these trade-offs. Through this combination, it is possible to gain insights about a running RL system efficiently and unobtrusively based on data stream monitoring and to create abstract representations that allow reasoning about the historical behaviour of the RL system. The obtained knowledge is exploited to provide feedback to the RL system for optimising its hyperparameters while making effective use of parallel resources. We introduce a novel history-aware epsilon-greedy logic for hyperparameter optimisation that instead of using static hyperparameters that are kept fixed for the whole training, adjusts the hyperparameters at runtime based on the analysis of the agent's performance over time windows in a single agent's lifetime. We tested the proposed approach in a 5G mobile communications case study that uses DQN, a variant of RL, for its decision-making. Our experiments demonstrated the effects of hyperparameter tuning using history on training stability and reward values. The encouraging results show that the proposed history-aware framework significantly improved performance compared to traditional hyperparameter tuning approaches.
☆ Inversion dynamics of class manifolds in deep learning reveals tradeoffs underlying generalisation
To achieve near-zero training error in a classification problem, the layers of a deep network have to disentangle the manifolds of data points with different labels, to facilitate the discrimination. However, excessive class separation can bring to overfitting since good generalisation requires learning invariant features, which involve some level of entanglement. We report on numerical experiments showing how the optimisation dynamics finds representations that balance these opposing tendencies with a non-monotonic trend. After a fast segregation phase, a slower rearrangement (conserved across data sets and architectures) increases the class entanglement. The training error at the inversion is remarkably stable under subsampling, and across network initialisations and optimisers, which characterises it as a property solely of the data structure and (very weakly) of the architecture. The inversion is the manifestation of tradeoffs elicited by well-defined and maximally stable elements of the training set, coined "stragglers", particularly influential for generalisation.
☆ Aux-Drop: Handling Haphazard Inputs in Online Learning Using Auxiliary Dropouts
Many real-world applications based on online learning produce streaming data that is haphazard in nature, i.e., contains missing features, features becoming obsolete in time, the appearance of new features at later points in time and a lack of clarity on the total number of input features. These challenges make it hard to build a learnable system for such applications, and almost no work exists in deep learning that addresses this issue. In this paper, we present Aux-Drop, an auxiliary dropout regularization strategy for online learning that handles the haphazard input features in an effective manner. Aux-Drop adapts the conventional dropout regularization scheme for the haphazard input feature space ensuring that the final output is minimally impacted by the chaotic appearance of such features. It helps to prevent the co-adaptation of especially the auxiliary and base features, as well as reduces the strong dependence of the output on any of the auxiliary inputs of the model. This helps in better learning for scenarios where certain features disappear in time or when new features are to be modeled. The efficacy of Aux-Drop has been demonstrated through extensive numerical experiments on SOTA benchmarking datasets that include Italy Power Demand, HIGGS, SUSY and multiple UCI datasets.
☆ Provable Data Subset Selection For Efficient Neural Network Training
Radial basis function neural networks (\emph{RBFNN}) are {well-known} for their capability to approximate any continuous function on a closed bounded set with arbitrary precision given enough hidden neurons. In this paper, we introduce the first algorithm to construct coresets for \emph{RBFNNs}, i.e., small weighted subsets that approximate the loss of the input data on any radial basis function network and thus approximate any function defined by an \emph{RBFNN} on the larger input data. In particular, we construct coresets for radial basis and Laplacian loss functions. We then use our coresets to obtain a provable data subset selection algorithm for training deep neural networks. Since our coresets approximate every function, they also approximate the gradient of each weight in a neural network, which is a particular function on the input. We then perform empirical evaluations on function approximation and dataset subset selection on popular network architectures and data sets, demonstrating the efficacy and accuracy of our coreset construction.
☆ ESCL: Equivariant Self-Contrastive Learning for Sentence Representations ICASSP 2023
Previous contrastive learning methods for sentence representations often focus on insensitive transformations to produce positive pairs, but neglect the role of sensitive transformations that are harmful to semantic representations. Therefore, we propose an Equivariant Self-Contrastive Learning (ESCL) method to make full use of sensitive transformations, which encourages the learned representations to be sensitive to certain types of transformations with an additional equivariant learning task. Meanwhile, in order to improve practicability and generality, ESCL simplifies the implementations of traditional equivariant contrastive methods to share model parameters from the perspective of multi-task learning. We evaluate our ESCL on semantic textual similarity tasks. The proposed method achieves better results while using fewer learning parameters compared to previous methods.
comment: accepted by ICASSP 2023
☆ The joint node degree distribution in the Erdős-Rényi network
The Erd\H{o}s-R\'enyi random graph is the simplest model for node degree distribution, and it is one of the most widely studied. In this model, pairs of $n$ vertices are selected and connected uniformly at random with probability $p$, consequently, the degrees for a given vertex follow the binomial distribution. If the number of vertices is large, the binomial can be approximated by Normal using the Central Limit Theorem, which is often allowed when $\min (np, n(1-p)) > 5$. This is true for every node independently. However, due to the fact that the degrees of nodes in a graph are not independent, we aim in this paper to test whether the degrees of per node collectively in the Erd\H{o}s-R\'enyi graph have a multivariate normal distribution MVN. A chi square goodness of fit test for the hypothesis that binomial is a distribution for the whole set of nodes is rejected because of the dependence between degrees. Before testing MVN we show that the covariance and correlation between the degrees of any pair of nodes in the graph are $p(1-p)$ and $1/(n-1)$, respectively. We test MVN considering two assumptions: independent and dependent degrees, and we obtain our results based on the percentages of rejected statistics of chi square, the $p$-values of Anderson Darling test, and a CDF comparison. We always achieve a good fit of multivariate normal distribution with large values of $n$ and $p$, and very poor fit when $n$ or $p$ are very small. The approximation seems valid when $np \geq 10$. We also compare the maximum likelihood estimate of $p$ in MVN distribution where we assume independence and dependence. The estimators are assessed using bias, variance and mean square error.
comment: 7 pages
☆ Entropic Wasserstein Component Analysis
Dimension reduction (DR) methods provide systematic approaches for analyzing high-dimensional data. A key requirement for DR is to incorporate global dependencies among original and embedded samples while preserving clusters in the embedding space. To achieve this, we combine the principles of optimal transport (OT) and principal component analysis (PCA). Our method seeks the best linear subspace that minimizes reconstruction error using entropic OT, which naturally encodes the neighborhood information of the samples. From an algorithmic standpoint, we propose an efficient block-majorization-minimization solver over the Stiefel manifold. Our experimental results demonstrate that our approach can effectively preserve high-dimensional clusters, leading to more interpretable and effective embeddings. Python code of the algorithms and experiments is available online.
☆ SLCA: Slow Learner with Classifier Alignment for Continual Learning on a Pre-trained Model
The goal of continual learning is to improve the performance of recognition models in learning sequentially arrived data. Although most existing works are established on the premise of learning from scratch, growing efforts have been devoted to incorporating the benefits of pre-training. However, how to adaptively exploit the pre-trained knowledge for each incremental task while maintaining its generalizability remains an open question. In this work, we present an extensive analysis for continual learning on a pre-trained model (CLPM), and attribute the key challenge to a progressive overfitting problem. Observing that selectively reducing the learning rate can almost resolve this issue in the representation layer, we propose a simple but extremely effective approach named Slow Learner with Classifier Alignment (SLCA), which further improves the classification layer by modeling the class-wise distributions and aligning the classification layers in a post-hoc fashion. Across a variety of scenarios, our proposal provides substantial improvements for CLPM (e.g., up to 49.76%, 50.05%, 44.69% and 40.16% on Split CIFAR-100, Split ImageNet-R, Split CUB-200 and Split Cars-196, respectively), and thus outperforms state-of-the-art approaches by a large margin. Based on such a strong baseline, critical factors and promising directions are analyzed in-depth to facilitate subsequent research.
comment: Tech report. 11 pages, 8 figures
☆ Segmentation method for cerebral blood vessels from MRA using hysteresis
Segmentation of cerebral blood vessels from Magnetic Resonance Imaging (MRI) is an open problem that could be solved with deep learning (DL). However, annotated data for training is often scarce. Due to the absence of open-source tools, we aim to develop a classical segmentation method that generates vessel ground truth from Magnetic Resonance Angiography for DL training of segmentation across a variety of modalities. The method combines size-specific Hessian filters, hysteresis thresholding and connected component correction. The optimal choice of processing steps was evaluated with a blinded scoring by a clinician using 24 3D images. The results show that all method steps are necessary to produce the highest (14.2/15) vessel segmentation quality score. Omitting the connected component correction caused the largest quality loss. The method, which is available on GitHub, can be used to train DL models for vessel segmentation.
☆ StyleDiff: Attribute Comparison Between Unlabeled Datasets in Latent Disentangled Space
One major challenge in machine learning applications is coping with mismatches between the datasets used in the development and those obtained in real-world applications. These mismatches may lead to inaccurate predictions and errors, resulting in poor product quality and unreliable systems. In this study, we propose StyleDiff to inform developers of the differences between the two datasets for the steady development of machine learning systems. Using disentangled image spaces obtained from recently proposed generative models, StyleDiff compares the two datasets by focusing on attributes in the images and provides an easy-to-understand analysis of the differences between the datasets. The proposed StyleDiff performs in $O (d N\log N)$, where $N$ is the size of the datasets and $d$ is the number of attributes, enabling the application to large datasets. We demonstrate that StyleDiff accurately detects differences between datasets and presents them in an understandable format using, for example, driving scenes datasets.
comment: 23 pages, 16 figures, under review
☆ Scalable Stochastic Gradient Riemannian Langevin Dynamics in Non-Diagonal Metrics
Bayesian neural network inference is often carried out using stochastic gradient sampling methods. For best performance the methods should use a Riemannian metric that improves posterior exploration by accounting for the local curvature, but the existing methods resort to simple diagonal metrics to remain computationally efficient. This loses some of the gains. We propose two non-diagonal metrics that can be used in stochastic samplers to improve convergence and exploration but that have only a minor computational overhead over diagonal metrics. We show that for neural networks with complex posteriors, caused e.g. by use of sparsity-inducing priors, using these metrics provides clear improvements. For some other choices the posterior is sufficiently easy also for the simpler metrics.
☆ Optimizing Sparse Linear Algebra Through Automatic Format Selection and Machine Learning
Sparse matrices are an integral part of scientific simulations. As hardware evolves new sparse matrix storage formats are proposed aiming to exploit optimizations specific to the new hardware. In the era of heterogeneous computing, users often are required to use multiple formats for their applications to remain optimal across the different available hardware, resulting in larger development times and maintenance overhead. A potential solution to this problem is the use of a lightweight auto-tuner driven by Machine Learning (ML) that would select for the user an optimal format from a pool of available formats that will match the characteristics of the sparsity pattern, target hardware and operation to execute. In this paper, we introduce Morpheus-Oracle, a library that provides a lightweight ML auto-tuner capable of accurately predicting the optimal format across multiple backends, targeting the major HPC architectures aiming to eliminate any format selection input by the end-user. From more than 2000 real-life matrices, we achieve an average classification accuracy and balanced accuracy of 92.63% and 80.22% respectively across the available systems. The adoption of the auto-tuner results in average speedup of 1.1x on CPUs and 1.5x to 8x on NVIDIA and AMD GPUs, with maximum speedups reaching up to 7x and 1000x respectively.
☆ Reward Informed Dreamer for Task Generalization in Reinforcement Learning
A long-standing goal of reinforcement learning is that algorithms can learn on training tasks and generalize well on unseen tasks like humans, where different tasks share similar dynamic with different reward functions. A general challenge is that it is nontrivial to quantitatively measure the similarities between these different tasks, which is vital for analyzing the task distribution and further designing algorithms with stronger generalization. To address this, we present a novel metric named Task Distribution Relevance (TDR) via optimal Q functions to capture the relevance of the task distribution quantitatively. In the case of tasks with a high TDR, i.e., the tasks differ significantly, we demonstrate that the Markovian policies cannot distinguish them, yielding poor performance accordingly. Based on this observation, we propose a framework of Reward Informed Dreamer (RID) with reward-informed world models, which captures invariant latent features over tasks and encodes reward signals into policies for distinguishing different tasks. In RID, we calculate the corresponding variational lower bound of the log-likelihood on the data, which includes a novel term to distinguish different tasks via states, based on reward-informed world models. Finally, extensive experiments in DeepMind control suite demonstrate that RID can significantly improve the performance of handling different tasks at the same time, especially for those with high TDR, and further generalize to unseen tasks effectively.
☆ Identification of Systematic Errors of Image Classifiers on Rare Subgroups
Despite excellent average-case performance of many image classifiers, their performance can substantially deteriorate on semantically coherent subgroups of the data that were under-represented in the training data. These systematic errors can impact both fairness for demographic minority groups as well as robustness and safety under domain shift. A major challenge is to identify such subgroups with subpar performance when the subgroups are not annotated and their occurrence is very rare. We leverage recent advances in text-to-image models and search in the space of textual descriptions of subgroups ("prompts") for subgroups where the target model has low performance on the prompt-conditioned synthesized data. To tackle the exponentially growing number of subgroups, we employ combinatorial testing. We denote this procedure as PromptAttack as it can be interpreted as an adversarial attack in a prompt space. We study subgroup coverage and identifiability with PromptAttack in a controlled setting and find that it identifies systematic errors with high accuracy. Thereupon, we apply PromptAttack to ImageNet classifiers and identify novel systematic errors on rare subgroups.
☆ Conceptual Reinforcement Learning for Language-Conditioned Tasks AAAI 2023
Despite the broad application of deep reinforcement learning (RL), transferring and adapting the policy to unseen but similar environments is still a significant challenge. Recently, the language-conditioned policy is proposed to facilitate policy transfer through learning the joint representation of observation and text that catches the compact and invariant information across environments. Existing studies of language-conditioned RL methods often learn the joint representation as a simple latent layer for the given instances (episode-specific observation and text), which inevitably includes noisy or irrelevant information and cause spurious correlations that are dependent on instances, thus hurting generalization performance and training efficiency. To address this issue, we propose a conceptual reinforcement learning (CRL) framework to learn the concept-like joint representation for language-conditioned policy. The key insight is that concepts are compact and invariant representations in human cognition through extracting similarities from numerous instances in real-world. In CRL, we propose a multi-level attention encoder and two mutual information constraints for learning compact and invariant concepts. Verified in two challenging environments, RTFM and Messenger, CRL significantly improves the training efficiency (up to 70%) and generalization ability (up to 30%) to the new environment dynamics.
comment: Accepted by AAAI 2023
☆ Semi-Federated Learning for Collaborative Intelligence in Massive IoT Networks
Implementing existing federated learning in massive Internet of Things (IoT) networks faces critical challenges such as imbalanced and statistically heterogeneous data and device diversity. To this end, we propose a semi-federated learning (SemiFL) framework to provide a potential solution for the realization of intelligent IoT. By seamlessly integrating the centralized and federated paradigms, our SemiFL framework shows high scalability in terms of the number of IoT devices even in the presence of computing-limited sensors. Furthermore, compared to traditional learning approaches, the proposed SemiFL can make better use of distributed data and computing resources, due to the collaborative model training between the edge server and local devices. Simulation results show the effectiveness of our SemiFL framework for massive IoT networks. The code can be found at https://github.com/niwanli/SemiFL_IoT.
comment: The paper has been accepted for publication in the IEEE Internet of Things Journal
☆ Invertible Kernel PCA with Random Fourier Features
Kernel principal component analysis (kPCA) is a widely studied method to construct a low-dimensional data representation after a nonlinear transformation. The prevailing method to reconstruct the original input signal from kPCA -- an important task for denoising -- requires us to solve a supervised learning problem. In this paper, we present an alternative method where the reconstruction follows naturally from the compression step. We first approximate the kernel with random Fourier features. Then, we exploit the fact that the nonlinear transformation is invertible in a certain subdomain. Hence, the name \emph{invertible kernel PCA (ikPCA)}. We experiment with different data modalities and show that ikPCA performs similarly to kPCA with supervised reconstruction on denoising tasks, making it a strong alternative.
comment: This work has been submitted to the IEEE for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessible
☆ Exploiting Contextual Structure to Generate Useful Auxiliary Tasks
Reinforcement learning requires interaction with an environment, which is expensive for robots. This constraint necessitates approaches that work with limited environmental interaction by maximizing the reuse of previous experiences. We propose an approach that maximizes experience reuse while learning to solve a given task by generating and simultaneously learning useful auxiliary tasks. To generate these tasks, we construct an abstract temporal logic representation of the given task and leverage large language models to generate context-aware object embeddings that facilitate object replacements. Counterfactual reasoning and off-policy methods allow us to simultaneously learn these auxiliary tasks while solving the given target task. We combine these insights into a novel framework for multitask reinforcement learning and experimentally show that our generated auxiliary tasks share similar underlying exploration requirements as the given task, thereby maximizing the utility of directed exploration. Our approach allows agents to automatically learn additional useful policies without extra environment interaction.
☆ Gauges and Accelerated Optimization over Smooth and/or Strongly Convex Sets
We consider feasibility and constrained optimization problems defined over smooth and/or strongly convex sets. These notions mirror their popular function counterparts but are much less explored in the first-order optimization literature. We propose new scalable, projection-free, accelerated first-order methods in these settings. Our methods avoid linear optimization or projection oracles, only using cheap one-dimensional linesearches and normal vector computations. Despite this, we derive optimal accelerated convergence guarantees of $O(1/T)$ for strongly convex problems, $O(1/T^2)$ for smooth problems, and accelerated linear convergence given both. Our algorithms and analysis are based on novel characterizations of the Minkowski gauge of smooth and/or strongly convex sets, which may be of independent interest: although the gauge is neither smooth nor strongly convex, we show the gauge squared inherits any structure present in the set.
comment: 22pages (32pages with references and appendix)
☆ Out-of-distribution Detection with Implicit Outlier Transformation
Outlier exposure (OE) is powerful in out-of-distribution (OOD) detection, enhancing detection capability via model fine-tuning with surrogate OOD data. However, surrogate data typically deviate from test OOD data. Thus, the performance of OE, when facing unseen OOD data, can be weakened. To address this issue, we propose a novel OE-based approach that makes the model perform well for unseen OOD situations, even for unseen OOD cases. It leads to a min-max learning scheme -- searching to synthesize OOD data that leads to worst judgments and learning from such OOD data for uniform performance in OOD detection. In our realization, these worst OOD data are synthesized by transforming original surrogate ones. Specifically, the associated transform functions are learned implicitly based on our novel insight that model perturbation leads to data transformation. Our methodology offers an efficient way of synthesizing OOD data, which can further benefit the detection model, besides the surrogate OOD data. We conduct extensive experiments under various OOD detection setups, demonstrating the effectiveness of our method against its advanced counterparts.
☆ SSL^2: Self-Supervised Learning meets Semi-Supervised Learning: Multiple Sclerosis Segmentation in 7T-MRI from large-scale 3T-MRI
Automated segmentation of multiple sclerosis (MS) lesions from MRI scans is important to quantify disease progression. In recent years, convolutional neural networks (CNNs) have shown top performance for this task when a large amount of labeled data is available. However, the accuracy of CNNs suffers when dealing with few and/or sparsely labeled datasets. A potential solution is to leverage the information available in large public datasets in conjunction with a target dataset which only has limited labeled data. In this paper, we propose a training framework, SSL2 (self-supervised-semi-supervised), for multi-modality MS lesion segmentation with limited supervision. We adopt self-supervised learning to leverage the knowledge from large public 3T datasets to tackle the limitations of a small 7T target dataset. To leverage the information from unlabeled 7T data, we also evaluate state-of-the-art semi-supervised methods for other limited annotation settings, such as small labeled training size and sparse annotations. We use the shifted-window (Swin) transformer1 as our backbone network. The effectiveness of self-supervised and semi-supervised training strategies is evaluated in our in-house 7T MRI dataset. The results indicate that each strategy improves lesion segmentation for both limited training data size and for sparse labeling scenarios. The combined overall framework further improves the performance substantially compared to either of its components alone. Our proposed framework thus provides a promising solution for future data/label-hungry 7T MS studies.
comment: Accepted at the International Society for Optics and Photonics - Medical Imaging (SPIE-MI) 2023
☆ Phase transition for detecting a small community in a large network
How to detect a small community in a large network is an interesting problem, including clique detection as a special case, where a naive degree-based $\chi^2$-test was shown to be powerful in the presence of an Erd\H{o}s-Renyi background. Using Sinkhorn's theorem, we show that the signal captured by the $\chi^2$-test may be a modeling artifact, and it may disappear once we replace the Erd\H{o}s-Renyi model by a broader network model. We show that the recent SgnQ test is more appropriate for such a setting. The test is optimal in detecting communities with sizes comparable to the whole network, but has never been studied for our setting, which is substantially different and more challenging. Using a degree-corrected block model (DCBM), we establish phase transitions of this testing problem concerning the size of the small community and the edge densities in small and large communities. When the size of the small community is larger than $\sqrt{n}$, the SgnQ test is optimal for it attains the computational lower bound (CLB), the information lower bound for methods allowing polynomial computation time. When the size of the small community is smaller than $\sqrt{n}$, we establish the parameter regime where the SgnQ test has full power and make some conjectures of the CLB. We also study the classical information lower bound (LB) and show that there is always a gap between the CLB and LB in our range of interest.
☆ Improved Regret Bounds for Online Kernel Selection under Bandit Feedback
In this paper, we improve the regret bound for online kernel selection under bandit feedback. Previous algorithm enjoys a $O((\Vert f\Vert^2_{\mathcal{H}_i}+1)K^{\frac{1}{3}}T^{\frac{2}{3}})$ expected bound for Lipschitz loss functions. We prove two types of regret bounds improving the previous bound. For smooth loss functions, we propose an algorithm with a $O(U^{\frac{2}{3}}K^{-\frac{1}{3}}(\sum^K_{i=1}L_T(f^\ast_i))^{\frac{2}{3}})$ expected bound where $L_T(f^\ast_i)$ is the cumulative losses of optimal hypothesis in $\mathbb{H}_{i}=\{f\in\mathcal{H}_i:\Vert f\Vert_{\mathcal{H}_i}\leq U\}$. The data-dependent bound keeps the previous worst-case bound and is smaller if most of candidate kernels match well with the data. For Lipschitz loss functions, we propose an algorithm with a $O(U\sqrt{KT}\ln^{\frac{2}{3}}{T})$ expected bound asymptotically improving the previous bound. We apply the two algorithms to online kernel selection with time constraint and prove new regret bounds matching or improving the previous $O(\sqrt{T\ln{K}} +\Vert f\Vert^2_{\mathcal{H}_i}\max\{\sqrt{T},\frac{T}{\sqrt{\mathcal{R}}}\})$ expected bound where $\mathcal{R}$ is the time budget. Finally, we empirically verify our algorithms on online regression and classification tasks.
☆ Learning Representation for Anomaly Detection of Vehicle Trajectories
Predicting the future trajectories of surrounding vehicles based on their history trajectories is a critical task in autonomous driving. However, when small crafted perturbations are introduced to those history trajectories, the resulting anomalous (or adversarial) trajectories can significantly mislead the future trajectory prediction module of the ego vehicle, which may result in unsafe planning and even fatal accidents. Therefore, it is of great importance to detect such anomalous trajectories of the surrounding vehicles for system safety, but few works have addressed this issue. In this work, we propose two novel methods for learning effective and efficient representations for online anomaly detection of vehicle trajectories. Different from general time-series anomaly detection, anomalous vehicle trajectory detection deals with much richer contexts on the road and fewer observable patterns on the anomalous trajectories themselves. To address these challenges, our methods exploit contrastive learning techniques and trajectory semantics to capture the patterns underlying the driving scenarios for effective anomaly detection under supervised and unsupervised settings, respectively. We conduct extensive experiments to demonstrate that our supervised method based on contrastive learning and unsupervised method based on reconstruction with semantic latent space can significantly improve the performance of anomalous trajectory detection in their corresponding settings over various baseline methods. We also demonstrate our methods' generalization ability to detect unseen patterns of anomalies.
comment: 8 pages paper, in anomaly detection of vehicle trajectory
☆ Curvature-Sensitive Predictive Coding with Approximate Laplace Monte Carlo
Predictive coding (PC) accounts of perception now form one of the dominant computational theories of the brain, where they prescribe a general algorithm for inference and learning over hierarchical latent probabilistic models. Despite this, they have enjoyed little export to the broader field of machine learning, where comparative generative modelling techniques have flourished. In part, this has been due to the poor performance of models trained with PC when evaluated by both sample quality and marginal likelihood. By adopting the perspective of PC as a variational Bayes algorithm under the Laplace approximation, we identify the source of these deficits to lie in the exclusion of an associated Hessian term in the PC objective function, which would otherwise regularise the sharpness of the probability landscape and prevent over-certainty in the approximate posterior. To remedy this, we make three primary contributions: we begin by suggesting a simple Monte Carlo estimated evidence lower bound which relies on sampling from the Hessian-parameterised variational posterior. We then derive a novel block diagonal approximation to the full Hessian matrix that has lower memory requirements and favourable mathematical properties. Lastly, we present an algorithm that combines our method with standard PC to reduce memory complexity further. We evaluate models trained with our approach against the standard PC framework on image benchmark datasets. Our approach produces higher log-likelihoods and qualitatively better samples that more closely capture the diversity of the data-generating distribution.
☆ Blackwell's Approachability with Time-Dependent Outcome Functions and Dot Products. Application to the Big Match
Blackwell's approachability is a very general sequential decision framework where a Decision Maker obtains vector-valued outcomes, and aims at the convergence of the average outcome to a given "target" set. Blackwell gave a sufficient condition for the decision maker having a strategy guaranteeing such a convergence against an adversarial environment, as well as what we now call the Blackwell's algorithm, which then ensures convergence. Blackwell's approachability has since been applied to numerous problems, in online learning and game theory, in particular. We extend this framework by allowing the outcome function and the dot product to be time-dependent. We establish a general guarantee for the natural extension to this framework of Blackwell's algorithm. In the case where the target set is an orthant, we present a family of time-dependent dot products which yields different convergence speeds for each coordinate of the average outcome. We apply this framework to the Big Match (one of the most important toy examples of stochastic games) where an $\epsilon$-uniformly optimal strategy for Player I is given by Blackwell's algorithm in a well-chosen auxiliary approachability problem.
☆ KGNv2: Separating Scale and Pose Prediction for Keypoint-based 6-DoF Grasp Pose Synthesis on RGB-D input IROS2023
We propose a new 6-DoF grasp pose synthesis approach from 2D/2.5D input based on keypoints. Keypoint-based grasp detector from image input has demonstrated promising results in the previous study, where the additional visual information provided by color images compensates for the noisy depth perception. However, it relies heavily on accurately predicting the location of keypoints in the image space. In this paper, we devise a new grasp generation network that reduces the dependency on precise keypoint estimation. Given an RGB-D input, our network estimates both the grasp pose from keypoint detection as well as scale towards the camera. We further re-design the keypoint output space in order to mitigate the negative impact of keypoint prediction noise to Perspective-n-Point (PnP) algorithm. Experiments show that the proposed method outperforms the baseline by a large margin, validating the efficacy of our approach. Finally, despite trained on simple synthetic objects, our method demonstrate sim-to-real capacity by showing competitive results in real-world robot experiments.
comment: Submitted to IROS2023
☆ An Improved Data Augmentation Scheme for Model Predictive Control Policy Approximation
This paper considers the problem of data generation for MPC policy approximation. Learning an approximate MPC policy from expert demonstrations requires a large data set consisting of optimal state-action pairs, sampled across the feasible state space. Yet, the key challenge of efficiently generating the training samples has not been studied widely. Recently, a sensitivity-based data augmentation framework for MPC policy approximation was proposed, where the parametric sensitivities are exploited to cheaply generate several additional samples from a single offline MPC computation. The error due to augmenting the training data set with inexact samples was shown to increase with the size of the neighborhood around each sample used for data augmentation. Building upon this work, this letter paper presents an improved data augmentation scheme based on predictor-corrector steps that enforces a user-defined level of accuracy, and shows that the error bound of the augmented samples are independent of the size of the neighborhood used for data augmentation.
☆ Variance-aware robust reinforcement learning with linear function approximation with heavy-tailed rewards
This paper presents two algorithms, AdaOFUL and VARA, for online sequential decision-making in the presence of heavy-tailed rewards with only finite variances. For linear stochastic bandits, we address the issue of heavy-tailed rewards by modifying the adaptive Huber regression and proposing AdaOFUL. AdaOFUL achieves a state-of-the-art regret bound of $\widetilde{\mathcal{O}}\big(d\big(\sum_{t=1}^T \nu_{t}^2\big)^{1/2}+d\big)$ as if the rewards were uniformly bounded, where $\nu_{t}^2$ is the observed conditional variance of the reward at round $t$, $d$ is the feature dimension, and $\widetilde{\mathcal{O}}(\cdot)$ hides logarithmic dependence. Building upon AdaOFUL, we propose VARA for linear MDPs, which achieves a tighter variance-aware regret bound of $\widetilde{\mathcal{O}}(d\sqrt{H\mathcal{G}^*K})$. Here, $H$ is the length of episodes, $K$ is the number of episodes, and $\mathcal{G}^*$ is a smaller instance-dependent quantity that can be bounded by other instance-dependent quantities when additional structural conditions on the MDP are satisfied. Our regret bound is superior to the current state-of-the-art bounds in three ways: (1) it depends on a tighter instance-dependent quantity and has optimal dependence on $d$ and $H$, (2) we can obtain further instance-dependent bounds of $\mathcal{G}^*$ under additional structural conditions on the MDP, and (3) our regret bound is valid even when rewards have only finite variances, achieving a level of generality unmatched by previous works. Overall, our modified adaptive Huber regression algorithm may serve as a useful building block in the design of algorithms for online problems with heavy-tailed rewards.
comment: 23 page main text, 42 page appendix
☆ Learning the Wrong Lessons: Inserting Trojans During Knowledge Distillation ICLR 2023
In recent years, knowledge distillation has become a cornerstone of efficiently deployed machine learning, with labs and industries using knowledge distillation to train models that are inexpensive and resource-optimized. Trojan attacks have contemporaneously gained significant prominence, revealing fundamental vulnerabilities in deep learning models. Given the widespread use of knowledge distillation, in this work we seek to exploit the unlabelled data knowledge distillation process to embed Trojans in a student model without introducing conspicuous behavior in the teacher. We ultimately devise a Trojan attack that effectively reduces student accuracy, does not alter teacher performance, and is efficiently constructible in practice.
comment: ICLR 2023 Workshop on Backdoor Attacks and Defenses in Machine Learning
☆ Generalization analysis of an unfolding network for analysis-based Compressed Sensing
Unfolding networks have shown promising results in the Compressed Sensing (CS) field. Yet, the investigation of their generalization ability is still in its infancy. In this paper, we perform generalization analysis of a state-of-the-art ADMM-based unfolding network, which jointly learns a decoder for CS and a sparsifying redundant analysis operator. To this end, we first impose a structural constraint on the learnable sparsifier, which parametrizes the network's hypothesis class. For the latter, we estimate its Rademacher complexity. With this estimate in hand, we deliver generalization error bounds for the examined network. Finally, the validity of our theory is assessed and numerical comparisons to a state-of-the-art unfolding network are made, on synthetic and real-world datasets. Our experimental results demonstrate that our proposed framework complies with our theoretical findings and outperforms the baseline, consistently for all datasets.
☆ Exploration of the search space of Gaussian graphical models for paired data
We consider the problem of learning a Gaussian graphical model in the case where the observations come from two dependent groups sharing the same variables. We focus on a family of coloured Gaussian graphical models specifically suited for the paired data problem. Commonly, graphical models are ordered by the submodel relationship so that the search space is a lattice, called the model inclusion lattice. We introduce a novel order between models, named the twin order. We show that, embedded with this order, the model space is a lattice that, unlike the model inclusion lattice, is distributive. Furthermore, we provide the relevant rules for the computation of the neighbours of a model. The latter are more efficient than the same operations in the model inclusion lattice, and are then exploited to achieve a more efficient exploration of the search space. These results can be applied to improve the efficiency of both greedy and Bayesian model search procedures. Here we implement a stepwise backward elimination procedure and evaluate its performance by means of simulations. Finally, the procedure is applied to learn a brain network from fMRI data where the two groups correspond to the left and right hemispheres, respectively.
☆ Optimal active particle navigation meets machine learning
The question of how "smart" active agents, like insects, microorganisms, or future colloidal robots need to steer to optimally reach or discover a target, such as an odor source, food, or a cancer cell in a complex environment has recently attracted great interest. Here, we provide an overview of recent developments, regarding such optimal navigation problems, from the micro- to the macroscale, and give a perspective by discussing some of the challenges which are ahead of us. Besides exemplifying an elementary approach to optimal navigation problems, the article focuses on works utilizing machine learning-based methods. Such learning-based approaches can uncover highly efficient navigation strategies even for problems that involve e.g. chaotic, high-dimensional, or unknown environments and are hardly solvable based on conventional analytical or simulation methods.
comment: 7 pages, 3 figures
☆ EfficientTempNet: Temporal Super-Resolution of Radar Rainfall ICLR 2023
Rainfall data collected by various remote sensing instruments such as radars or satellites has different space-time resolutions. This study aims to improve the temporal resolution of radar rainfall products to help with more accurate climate change modeling and studies. In this direction, we introduce a solution based on EfficientNetV2, namely EfficientTempNet, to increase the temporal resolution of radar-based rainfall products from 10 minutes to 5 minutes. We tested EfficientRainNet over a dataset for the state of Iowa, US, and compared its performance to three different baselines to show that EfficientTempNet presents a viable option for better climate change monitoring.
comment: Published as a workshop paper at Tackling Climate Change with Machine Learning, ICLR 2023
☆ Position Paper on Dataset Engineering to Accelerate Science AAAI
Data is a critical element in any discovery process. In the last decades, we observed exponential growth in the volume of available data and the technology to manipulate it. However, data is only practical when one can structure it for a well-defined task. For instance, we need a corpus of text broken into sentences to train a natural language machine-learning model. In this work, we will use the token \textit{dataset} to designate a structured set of data built to perform a well-defined task. Moreover, the dataset will be used in most cases as a blueprint of an entity that at any moment can be stored as a table. Specifically, in science, each area has unique forms to organize, gather and handle its datasets. We believe that datasets must be a first-class entity in any knowledge-intensive process, and all workflows should have exceptional attention to datasets' lifecycle, from their gathering to uses and evolution. We advocate that science and engineering discovery processes are extreme instances of the need for such organization on datasets, claiming for new approaches and tooling. Furthermore, these requirements are more evident when the discovery workflow uses artificial intelligence methods to empower the subject-matter expert. In this work, we discuss an approach to bringing datasets as a critical entity in the discovery process in science. We illustrate some concepts using material discovery as a use case. We chose this domain because it leverages many significant problems that can be generalized to other science fields.
comment: Published at 2nd Annual AAAI Workshop on AI to Accelerate Science and Engineering (AI2ASE) https://ai-2-ase.github.io/papers/16%5cSubmission%5cAAAI_Dataset_Engineering-8.pdf
♻ ☆ Efficient Online Reinforcement Learning with Offline Data
Sample efficiency and exploration remain major challenges in online reinforcement learning (RL). A powerful approach that can be applied to address these issues is the inclusion of offline data, such as prior trajectories from a human expert or a sub-optimal exploration policy. Previous methods have relied on extensive modifications and additional complexity to ensure the effective use of this data. Instead, we ask: can we simply apply existing off-policy methods to leverage offline data when learning online? In this work, we demonstrate that the answer is yes; however, a set of minimal but important changes to existing off-policy RL algorithms are required to achieve reliable performance. We extensively ablate these design choices, demonstrating the key factors that most affect performance, and arrive at a set of recommendations that practitioners can readily apply, whether their data comprise a small number of expert demonstrations or large volumes of sub-optimal trajectories. We see that correct application of these simple recommendations can provide a $\mathbf{2.5\times}$ improvement over existing approaches across a diverse set of competitive benchmarks, with no additional computational overhead.
comment: To reproduce our results and use our codebase, see https://github.com/ikostrikov/rlpd
♻ ☆ Bayesian Weapon System Reliability Modeling with Cox-Weibull Neural Network
We propose to integrate weapon system features (such as weapon system manufacturer, deployment time and location, storage time and location, etc.) into a parameterized Cox-Weibull [1] reliability model via a neural network, like DeepSurv [2], to improve predictive maintenance. In parallel, we develop an alternative Bayesian model by parameterizing the Weibull parameters with a neural network and employing dropout methods such as Monte-Carlo (MC)-dropout for comparative purposes. Due to data collection procedures in weapon system testing we employ a novel interval-censored log-likelihood which incorporates Monte-Carlo Markov Chain (MCMC) [3] sampling of the Weibull parameters during gradient descent optimization. We compare classification metrics such as receiver operator curve (ROC) area under the curve (AUC), precision-recall (PR) AUC, and F scores to show our model generally outperforms traditional powerful models such as XGBoost and the current standard conditional Weibull probability density estimation model.
comment: Pre-print with minor revisions, published at The 69th Annual Reliability and Maintainability Symposium, January 23-26, 2023, FL, USA
♻ ☆ Designing Universal Causal Deep Learning Models: The Geometric (Hyper)Transformer
Several problems in stochastic analysis are defined through their geometry, and preserving that geometric structure is essential to generating meaningful predictions. Nevertheless, how to design principled deep learning (DL) models capable of encoding these geometric structures remains largely unknown. We address this open problem by introducing a universal causal geometric DL framework in which the user specifies a suitable pair of metric spaces $\mathscr{X}$ and $\mathscr{Y}$ and our framework returns a DL model capable of causally approximating any ``regular'' map sending time series in $\mathscr{X}^{\mathbb{Z}}$ to time series in $\mathscr{Y}^{\mathbb{Z}}$ while respecting their forward flow of information throughout time. Suitable geometries on $\mathscr{Y}$ include various (adapted) Wasserstein spaces arising in optimal stopping problems, a variety of statistical manifolds describing the conditional distribution of continuous-time finite state Markov chains, and all Fr\'{e}chet spaces admitting a Schauder basis, e.g. as in classical finance. Suitable spaces $\mathscr{X}$ are compact subsets of any Euclidean space. Our results all quantitatively express the number of parameters needed for our DL model to achieve a given approximation error as a function of the target map's regularity and the geometric structure both of $\mathscr{X}$ and of $\mathscr{Y}$. Even when omitting any temporal structure, our universal approximation theorems are the first guarantees that H\"{o}lder functions, defined between such $\mathscr{X}$ and $\mathscr{Y}$ can be approximated by DL models.
comment: Main Body: 31 Pages, Proofs: 16 Pages, Figures: 13, Tables: 3
♻ ☆ Provably Safe Reinforcement Learning with Step-wise Violation Constraints
In this paper, we investigate a novel safe reinforcement learning problem with step-wise violation constraints. Our problem differs from existing works in that we consider stricter step-wise violation constraints and do not assume the existence of safe actions, making our formulation more suitable for safety-critical applications which need to ensure safety in all decision steps and may not always possess safe actions, e.g., robot control and autonomous driving. We propose a novel algorithm SUCBVI, which guarantees $\widetilde{O}(\sqrt{ST})$ step-wise violation and $\widetilde{O}(\sqrt{H^3SAT})$ regret. Lower bounds are provided to validate the optimality in both violation and regret performance with respect to $S$ and $T$. Moreover, we further study a novel safe reward-free exploration problem with step-wise violation constraints. For this problem, we design an $(\varepsilon,\delta)$-PAC algorithm SRF-UCRL, which achieves nearly state-of-the-art sample complexity $\widetilde{O}((\frac{S^2AH^2}{\varepsilon}+\frac{H^4SA}{\varepsilon^2})(\log(\frac{1}{\delta})+S))$, and guarantees $\widetilde{O}(\sqrt{ST})$ violation during the exploration. The experimental results demonstrate the superiority of our algorithms in safety performance, and corroborate our theoretical results.
♻ ☆ Matching Map Recovery with an Unknown Number of Outliers AISTATS 2023
We consider the problem of finding the matching map between two sets of $d$-dimensional noisy feature-vectors. The distinctive feature of our setting is that we do not assume that all the vectors of the first set have their corresponding vector in the second set. If $n$ and $m$ are the sizes of these two sets, we assume that the matching map that should be recovered is defined on a subset of unknown cardinality $k^*\le \min(n,m)$. We show that, in the high-dimensional setting, if the signal-to-noise ratio is larger than $5(d\log(4nm/\alpha))^{1/4}$, then the true matching map can be recovered with probability $1-\alpha$. Interestingly, this threshold does not depend on $k^*$ and is the same as the one obtained in prior work in the case of $k = \min(n,m)$. The procedure for which the aforementioned property is proved is obtained by a data-driven selection among candidate mappings $\{\hat\pi_k:k\in[\min(n,m)]\}$. Each $\hat\pi_k$ minimizes the sum of squares of distances between two sets of size $k$. The resulting optimization problem can be formulated as a minimum-cost flow problem, and thus solved efficiently. Finally, we report the results of numerical experiments on both synthetic and real-world data that illustrate our theoretical results and provide further insight into the properties of the algorithms studied in this work.
comment: 16 pages, 8 figures, 1 table; AISTATS 2023
♻ ☆ Complex QA and language models hybrid architectures, Survey
This paper reviews the state-of-the-art of hybrid language models architectures and strategies for "complex" question-answering (QA, CQA, CPS). Large Language Models (LLM) are good at leveraging public data on standard problems but once you want to tackle more specific complex questions or problems you may need specific architecture, knowledge, skills, methods, sensitive data protection, explainability, human approval and versatile feedback... We identify key elements augmenting LLM to solve complex questions or problems. We extend findings from the robust community edited research papers BIG, BLOOM and HELM which open source, benchmark and analyze limits and challenges of LLM in terms of tasks complexity and strict evaluation on accuracy (e.g. fairness, robustness, toxicity, ...). Recent projects like ChatGPT and GALACTICA have allowed non-specialists to grasp the great potential as well as the equally strong limitations of language models in complex QA. Hybridizing these models with different components could allow to overcome these different limits and go much further. We discuss some challenges associated with complex QA, including domain adaptation, decomposition and efficient multi-step QA, long form and non-factoid QA, safety and multi-sensitivity data protection, multimodal search, hallucinations, explainability and truthfulness, temproal reasoning. Therefore, we analyze current solutions and promising research trends, using elements such as: hybrid LLM architectures, active human reinforcement learning supervised with AI, prompting adaptation, neuro-symbolic and structured knowledge grounding, program synthesis, iterated decomposition and others.
♻ ☆ ELODIN: Naming Concepts in Embedding Spaces
Despite recent advancements, the field of text-to-image synthesis still suffers from lack of fine-grained control. Using only text, it remains challenging to deal with issues such as concept coherence and concept contamination. We propose a method to enhance control by generating specific concepts that can be reused throughout multiple images, effectively expanding natural language with new words that can be combined much like a painter's palette. Unlike previous contributions, our method does not copy visuals from input data and can generate concepts through text alone. We perform a set of comparisons that finds our method to be a significant improvement over text-only prompts.
comment: Added quantitative data, fixed formatting issues
♻ ☆ Three New Validators and a Large-Scale Benchmark Ranking for Unsupervised Domain Adaptation
Changes to hyperparameters can have a dramatic effect on model accuracy. Thus, the tuning of hyperparameters plays an important role in optimizing machine-learning models. An integral part of the hyperparameter-tuning process is the evaluation of model checkpoints, which is done through the use of "validators". In a supervised setting, these validators evaluate checkpoints by computing accuracy on a validation set that has labels. In contrast, in an unsupervised setting, the validation set has no such labels. Without any labels, it is impossible to compute accuracy, so validators must estimate accuracy instead. But what is the best approach to estimating accuracy? In this paper, we consider this question in the context of unsupervised domain adaptation (UDA). Specifically, we propose three new validators, and we compare and rank them against five other existing validators, on a large dataset of 1,000,000 checkpoints. Extensive experimental results show that two of our proposed validators achieve state-of-the-art performance in various settings. Finally, we find that in many cases, the state-of-the-art is obtained by a simple baseline method. To the best of our knowledge, this is the largest empirical study of UDA validators to date. Code is available at https://www.github.com/KevinMusgrave/powerful-benchmarker.
comment: This paper was previously titled Benchmarking Validation Methods for Unsupervised Domain Adaptation. This version contains new experiments, analysis, and figures
♻ ☆ Synthesizer Preset Interpolation using Transformer Auto-Encoders ICASSP 2023
Sound synthesizers are widespread in modern music production but they increasingly require expert skills to be mastered. This work focuses on interpolation between presets, i.e., sets of values of all sound synthesis parameters, to enable the intuitive creation of new sounds from existing ones. We introduce a bimodal auto-encoder neural network, which simultaneously processes presets using multi-head attention blocks, and audio using convolutions. This model has been tested on a popular frequency modulation synthesizer with more than one hundred parameters. Experiments have compared the model to related architectures and methods, and have demonstrated that it performs smoother interpolations. After training, the proposed model can be integrated into commercial synthesizers for live interpolation or sound design tasks.
comment: Accepted to IEEE ICASSP 2023
♻ ☆ Deep network series for large-scale high-dynamic range imaging
We propose a new approach for large-scale high-dynamic range computational imaging. Deep Neural Networks (DNNs) trained end-to-end can solve linear inverse imaging problems almost instantaneously. While unfolded architectures provide robustness to measurement setting variations, embedding large-scale measurement operators in DNN architectures is impractical. Alternative Plug-and-Play (PnP) approaches, where the denoising DNNs are blind to the measurement setting, have proven effective to address scalability and high-dynamic range challenges, but rely on highly iterative algorithms. We propose a residual DNN series approach, also interpretable as a learned version of matching pursuit, where the reconstructed image is a sum of residual images progressively increasing the dynamic range, and estimated iteratively by DNNs taking the back-projected data residual of the previous iteration as input. We demonstrate on radio-astronomical imaging simulations that a series of only few terms provides a reconstruction quality competitive with PnP, at a fraction of the cost.
comment: 5 pages, 4 figures, 1 table
♻ ☆ LidarCLIP or: How I Learned to Talk to Point Clouds
Research connecting text and images has recently seen several breakthroughs, with models like CLIP, DALL-E 2, and Stable Diffusion. However, the connection between text and other visual modalities, such as lidar data, has received less attention, prohibited by the lack of text-lidar datasets. In this work, we propose LidarCLIP, a mapping from automotive point clouds to a pre-existing CLIP embedding space. Using image-lidar pairs, we supervise a point cloud encoder with the image CLIP embeddings, effectively relating text and lidar data with the image domain as an intermediary. We show the effectiveness of LidarCLIP by demonstrating that lidar-based retrieval is generally on par with image-based retrieval, but with complementary strengths and weaknesses. By combining image and lidar features, we improve upon both single-modality methods and enable a targeted search for challenging detection scenarios under adverse sensor conditions. We also explore zero-shot classification and show that LidarCLIP outperforms existing attempts to use CLIP for point clouds by a large margin. Finally, we leverage our compatibility with CLIP to explore a range of applications, such as point cloud captioning and lidar-to-image generation, without any additional training. Code and pre-trained models are available at https://github.com/atonderski/lidarclip.
♻ ☆ The Effect of Modeling Human Rationality Level on Learning Rewards from Multiple Feedback Types AAAI 2023
When inferring reward functions from human behavior (be it demonstrations, comparisons, physical corrections, or e-stops), it has proven useful to model the human as making noisy-rational choices, with a "rationality coefficient" capturing how much noise or entropy we expect to see in the human behavior. Prior work typically sets the rationality level to a constant value, regardless of the type, or quality, of human feedback. However, in many settings, giving one type of feedback (e.g. a demonstration) may be much more difficult than a different type of feedback (e.g. answering a comparison query). Thus, we expect to see more or less noise depending on the type of human feedback. In this work, we advocate that grounding the rationality coefficient in real data for each feedback type, rather than assuming a default value, has a significant positive effect on reward learning. We test this in both simulated experiments and in a user study with real human feedback. We find that overestimating human rationality can have dire effects on reward learning accuracy and regret. We also find that fitting the rationality coefficient to human data enables better reward learning, even when the human deviates significantly from the noisy-rational choice model due to systematic biases. Further, we find that the rationality level affects the informativeness of each feedback type: surprisingly, demonstrations are not always the most informative -- when the human acts very suboptimally, comparisons actually become more informative, even when the rationality level is the same for both. Ultimately, our results emphasize the importance and advantage of paying attention to the assumed human-rationality level, especially when agents actively learn from multiple types of human feedback.
comment: Published at AAAI 2023; 10 pages, 5 figures plus appendices
♻ ☆ Building Normalizing Flows with Stochastic Interpolants ICLR 2023
A generative model based on a continuous-time normalizing flow between any pair of base and target probability densities is proposed. The velocity field of this flow is inferred from the probability current of a time-dependent density that interpolates between the base and the target in finite time. Unlike conventional normalizing flow inference methods based the maximum likelihood principle, which require costly backpropagation through ODE solvers, our interpolant approach leads to a simple quadratic loss for the velocity itself which is expressed in terms of expectations that are readily amenable to empirical estimation. The flow can be used to generate samples from either the base or target, and to estimate the likelihood at any time along the interpolant. In addition, the flow can be optimized to minimize the path length of the interpolant density, thereby paving the way for building optimal transport maps. In situations where the base is a Gaussian density, we also show that the velocity of our normalizing flow can also be used to construct a diffusion model to sample the target as well as estimate its score. However, our approach shows that we can bypass this diffusion completely and work at the level of the probability flow with greater simplicity, opening an avenue for methods based solely on ordinary differential equations as an alternative to those based on stochastic differential equations. Benchmarking on density estimation tasks illustrates that the learned flow can match and surpass conventional continuous flows at a fraction of the cost, and compares well with diffusions on image generation on CIFAR-10 and ImageNet $32\times32$. The method scales ab-initio ODE flows to previously unreachable image resolutions, demonstrated up to $128\times128$.
comment: ICLR 2023
♻ ☆ Structured State Space Models for In-Context Reinforcement Learning
Structured state space sequence (S4) models have recently achieved state-of-the-art performance on long-range sequence modeling tasks. These models also have fast inference speeds and parallelisable training, making them potentially useful in many reinforcement learning settings. We propose a modification to a variant of S4 that enables us to initialise and reset the hidden state in parallel, allowing us to tackle reinforcement learning tasks. We show that our modified architecture runs asymptotically faster than Transformers and performs better than LSTM models on a simple memory-based task. Then, by leveraging the model's ability to handle long-range sequences, we achieve strong performance on a challenging meta-learning task in which the agent is given a randomly-sampled continuous control environment, combined with a randomly-sampled linear projection of the environment's observations and actions. Furthermore, we show the resulting model can adapt to out-of-distribution held-out tasks. Overall, the results presented in this paper suggest that the S4 models are a strong contender for the default architecture used for in-context reinforcement learning
♻ ☆ Masked Autoencoder for Self-Supervised Pre-training on Lidar Point Clouds
Masked autoencoding has become a successful pretraining paradigm for Transformer models for text, images, and, recently, point clouds. Raw automotive datasets are suitable candidates for self-supervised pre-training as they generally are cheap to collect compared to annotations for tasks like 3D object detection (OD). However, the development of masked autoencoders for point clouds has focused solely on synthetic and indoor data. Consequently, existing methods have tailored their representations and models toward small and dense point clouds with homogeneous point densities. In this work, we study masked autoencoding for point clouds in an automotive setting, which are sparse and for which the point density can vary drastically among objects in the same scene. To this end, we propose Voxel-MAE, a simple masked autoencoding pre-training scheme designed for voxel representations. We pre-train the backbone of a Transformer-based 3D object detector to reconstruct masked voxels and to distinguish between empty and non-empty voxels. Our method improves the 3D OD performance by 1.75 mAP points and 1.05 NDS on the challenging nuScenes dataset. Further, we show that by pre-training with Voxel-MAE, we require only 40% of the annotated data to outperform a randomly initialized equivalent. Code available at https://github.com/georghess/voxel-mae
♻ ☆ Hindsight States: Blending Sim and Real Task Elements for Efficient Reinforcement Learning
Reinforcement learning has shown great potential in solving complex tasks when large amounts of data can be generated with little effort. In robotics, one approach to generate training data builds on simulations based on dynamics models derived from first principles. However, for tasks that, for instance, involve complex soft robots, devising such models is substantially more challenging. Being able to train effectively in increasingly complicated scenarios with reinforcement learning enables to take advantage of complex systems such as soft robots. Here, we leverage the imbalance in complexity of the dynamics to learn more sample-efficiently. We (i) abstract the task into distinct components, (ii) off-load the simple dynamics parts into the simulation, and (iii) multiply these virtual parts to generate more data in hindsight. Our new method, Hindsight States (HiS), uses this data and selects the most useful transitions for training. It can be used with an arbitrary off-policy algorithm. We validate our method on several challenging simulated tasks and demonstrate that it improves learning both alone and when combined with an existing hindsight algorithm, Hindsight Experience Replay (HER). Finally, we evaluate HiS on a physical system and show that it boosts performance on a complex table tennis task with a muscular robot. Videos and code of the experiments can be found on webdav.tuebingen.mpg.de/his/.
♻ ☆ Linearly Mapping from Image to Text Space ICLR 2023
The extent to which text-only language models (LMs) learn to represent features of the non-linguistic world is an open question. Prior work has shown that pretrained LMs can be taught to caption images when a vision model's parameters are optimized to encode images in the language space. We test a stronger hypothesis: that the conceptual representations learned by frozen text-only models and vision-only models are similar enough that this can be achieved with a linear map. We show that the image representations from vision models can be transferred as continuous prompts to frozen LMs by training only a single linear projection. Using these to prompt the LM achieves competitive performance on captioning and visual question answering tasks compared to models that tune both the image encoder and text decoder (such as the MAGMA model). We compare three image encoders with increasing amounts of linguistic supervision seen during pretraining: BEIT (no linguistic information), NF-ResNET (lexical category information), and CLIP (full natural language descriptions). We find that all three encoders perform equally well at transferring visual property information to the language model (e.g., whether an animal is large or small), but that image encoders pretrained with linguistic supervision more saliently encode category information (e.g., distinguishing hippo vs. elephant) and thus perform significantly better on benchmark language-and-vision tasks. Our results indicate that LMs encode conceptual information structurally similarly to vision-based models, even those that are solely trained on images. Code is available here: https://github.com/jmerullo/limber
comment: Accepted at ICLR 2023
♻ ☆ Energy-Latency Attacks via Sponge Poisoning
Sponge examples are test-time inputs carefully optimized to increase energy consumption and latency of neural networks when deployed on hardware accelerators. In this work, we are the first to demonstrate that sponge examples can also be injected at training time, via an attack that we call sponge poisoning. This attack allows one to increase the energy consumption and latency of machine-learning models indiscriminately on each test-time input. We present a novel formalization for sponge poisoning, overcoming the limitations related to the optimization of test-time sponge examples, and show that this attack is possible even if the attacker only controls a few model updates; for instance, if model training is outsourced to an untrusted third-party or distributed via federated learning. Our extensive experimental analysis shows that sponge poisoning can almost completely vanish the effect of hardware accelerators. We also analyze the activations of poisoned models, identifying which components are more vulnerable to this attack. Finally, we examine the feasibility of countermeasures against sponge poisoning to decrease energy consumption, showing that sanitization methods may be overly expensive for most of the users.
comment: Preprint;16 pages
♻ ☆ Target Specific De Novo Design of Drug Candidate Molecules with Graph Transformer-based Generative Adversarial Networks
Discovering novel drug candidate molecules is one of the most fundamental and critical steps in drug development. Generative deep learning models, which create synthetic data given a probability distribution, have been developed with the purpose of picking completely new samples from a partially known space. Generative models offer high potential for designing de novo molecules; however, in order for them to be useful in real-life drug development pipelines, these models should be able to design target-specific molecules, which is the next step in this field. In this study, we propose DrugGEN, for the de novo design of drug candidate molecules that interact with selected target proteins. The proposed system represents compounds and protein structures as graphs and processes them via serially connected two generative adversarial networks comprising graph transformers. DrugGEN is trained using a large dataset of compounds from ChEMBL and target-specific bioactive molecules, to design effective and specific inhibitory molecules against the AKT1 protein, which has critical importance for developing treatments against various types of cancer. On fundamental benchmarks, DrugGEN models have either competitive or better performance against other methods. To assess the target-specific generation performance, we conducted further in silico analysis with molecular docking and deep learning-based bioactivity prediction. Results indicate that de novo molecules have high potential for interacting with the AKT1 protein structure in the level of its native ligand. DrugGEN can be used to design completely novel and effective target-specific drug candidate molecules for any druggable protein, given target features and a dataset of experimental bioactivities. Code base, datasets, results and trained models of DrugGEN are available at https://github.com/HUBioDataLab/DrugGEN
♻ ☆ Generalization Bounds via Information Density and Conditional Information Density
We present a general approach, based on exponential inequalities, to derive bounds on the generalization error of randomized learning algorithms. Using this approach, we provide bounds on the average generalization error as well as bounds on its tail probability, for both the PAC-Bayesian and single-draw scenarios. Specifically, for the case of sub-Gaussian loss functions, we obtain novel bounds that depend on the information density between the training data and the output hypothesis. When suitably weakened, these bounds recover many of the information-theoretic bounds available in the literature. We also extend the proposed exponential-inequality approach to the setting recently introduced by Steinke and Zakynthinou (2020), where the learning algorithm depends on a randomly selected subset of the available training data. For this setup, we present bounds for bounded loss functions in terms of the conditional information density between the output hypothesis and the random variable determining the subset choice, given all training data. Through our approach, we recover the average generalization bound presented by Steinke and Zakynthinou (2020) and extend it to the PAC-Bayesian and single-draw scenarios. For the single-draw scenario, we also obtain novel bounds in terms of the conditional $\alpha$-mutual information and the conditional maximal leakage.
comment: Published in Journal on Selected Areas in Information Theory (JSAIT). This version incorporates a correction to the JSAIT version. The correction is detailed at https://gdurisi.github.io/files/2021/jsait-correction.pdf
♻ ☆ Enhancing Knowledge Graph Embedding Models with Semantic-driven Loss Functions
Knowledge graph embedding models (KGEMs) are used for various tasks related to knowledge graphs (KGs), including link prediction. They are trained with loss functions that are computed considering a batch of scored triples and their corresponding labels. Traditional approaches consider the label of a triple to be either true or false. However, recent works suggest that all negative triples should not be valued equally. In line with this recent assumption, we posit that semantically valid negative triples might be high-quality negative triples. As such, loss functions should treat them differently from semantically invalid negative ones. To this aim, we propose semantic-driven versions for the three main loss functions for link prediction. In particular, we treat the scores of negative triples differently by injecting background knowledge about relation domains and ranges into the loss functions. In an extensive and controlled experimental setting, we show that the proposed loss functions systematically provide satisfying results on three public benchmark KGs underpinned with different schemas, which demonstrates both the generality and superiority of our proposed approach. In fact, the proposed loss functions do (1) lead to better MRR and Hits@$10$ values, (2) drive KGEMs towards better semantic awareness. This highlights that semantic information globally improves KGEMs, and thus should be incorporated into loss functions. Domains and ranges of relations being largely available in schema-defined KGs, this makes our approach both beneficial and widely usable in practice.
♻ ☆ Generalized Balancing Weights via Deep Neural Networks
We present generalized balancing weights, Neural Balancing Weights (NBW), to estimate the causal effects for an arbitrary mixture of discrete and continuous interventions. The weights were obtained by directly estimating the density ratio between the source and balanced distributions by optimizing the variational representation of $f$-divergence. For this, we selected $\alpha$-divergence since it has good properties for optimization: It has an estimator whose sample complexity is independent of it's ground truth value and unbiased mini-batch gradients and is advantageous for the vanishing gradient problem. In addition, we provide a method for checking the balance of the distribution changed by the weights. If the balancing is imperfect, the weights can be improved by adding new balancing weights. Our method can be conveniently implemented with any present deep-learning libraries, and weights can be used in most state-of-the-art supervised algorithms. The code for our method is available online.
♻ ☆ Wild Patterns Reloaded: A Survey of Machine Learning Security against Training Data Poisoning
The success of machine learning is fueled by the increasing availability of computing power and large training datasets. The training data is used to learn new models or update existing ones, assuming that it is sufficiently representative of the data that will be encountered at test time. This assumption is challenged by the threat of poisoning, an attack that manipulates the training data to compromise the model's performance at test time. Although poisoning has been acknowledged as a relevant threat in industry applications, and a variety of different attacks and defenses have been proposed so far, a complete systematization and critical review of the field is still missing. In this survey, we provide a comprehensive systematization of poisoning attacks and defenses in machine learning, reviewing more than 100 papers published in the field in the last 15 years. We start by categorizing the current threat models and attacks, and then organize existing defenses accordingly. While we focus mostly on computer-vision applications, we argue that our systematization also encompasses state-of-the-art attacks and defenses for other data modalities. Finally, we discuss existing resources for research in poisoning, and shed light on the current limitations and open research questions in this research field.
comment: 35 pages, Accepted at ACM Computing Surveys
♻ ☆ SAM as an Optimal Relaxation of Bayes ICLR 2023
Sharpness-aware minimization (SAM) and related adversarial deep-learning methods can drastically improve generalization, but their underlying mechanisms are not yet fully understood. Here, we establish SAM as a relaxation of the Bayes objective where the expected negative-loss is replaced by the optimal convex lower bound, obtained by using the so-called Fenchel biconjugate. The connection enables a new Adam-like extension of SAM to automatically obtain reasonable uncertainty estimates, while sometimes also improving its accuracy. By connecting adversarial and Bayesian methods, our work opens a new path to robustness.
comment: Accepted at ICLR 2023
♻ ☆ MobileBrick: Building LEGO for 3D Reconstruction on Mobile Devices CVPR 2023
High-quality 3D ground-truth shapes are critical for 3D object reconstruction evaluation. However, it is difficult to create a replica of an object in reality, and even 3D reconstructions generated by 3D scanners have artefacts that cause biases in evaluation. To address this issue, we introduce a novel multi-view RGBD dataset captured using a mobile device, which includes highly precise 3D ground-truth annotations for 153 object models featuring a diverse set of 3D structures. We obtain precise 3D ground-truth shape without relying on high-end 3D scanners by utilising LEGO models with known geometry as the 3D structures for image capture. The distinct data modality offered by high-resolution RGB images and low-resolution depth maps captured on a mobile device, when combined with precise 3D geometry annotations, presents a unique opportunity for future research on high-fidelity 3D reconstruction. Furthermore, we evaluate a range of 3D reconstruction algorithms on the proposed dataset. Project page: http://code.active.vision/MobileBrick/
comment: To be appeared at CVPR 2023
♻ ☆ Certified Training: Small Boxes are All You Need ICLR23
To obtain, deterministic guarantees of adversarial robustness, specialized training methods are used. We propose, SABR, a novel such certified training method, based on the key insight that propagating interval bounds for a small but carefully selected subset of the adversarial input region is sufficient to approximate the worst-case loss over the whole region while significantly reducing approximation errors. We show in an extensive empirical evaluation that SABR outperforms existing certified defenses in terms of both standard and certifiable accuracies across perturbation magnitudes and datasets, pointing to a new class of certified training methods promising to alleviate the robustness-accuracy trade-off.
comment: Accepted at ICLR23 as Spotlight
♻ ☆ A view of mini-batch SGD via generating functions: conditions of convergence, phase transitions, benefit from negative momenta ICLR2023
Mini-batch SGD with momentum is a fundamental algorithm for learning large predictive models. In this paper we develop a new analytic framework to analyze noise-averaged properties of mini-batch SGD for linear models at constant learning rates, momenta and sizes of batches. Our key idea is to consider the dynamics of the second moments of model parameters for a special family of "Spectrally Expressible" approximations. This allows to obtain an explicit expression for the generating function of the sequence of loss values. By analyzing this generating function, we find, in particular, that 1) the SGD dynamics exhibits several convergent and divergent regimes depending on the spectral distributions of the problem; 2) the convergent regimes admit explicit stability conditions, and explicit loss asymptotics in the case of power-law spectral distributions; 3) the optimal convergence rate can be achieved at negative momenta. We verify our theoretical predictions by extensive experiments with MNIST, CIFAR10 and synthetic problems, and find a good quantitative agreement.
comment: The revised version accepted at ICLR2023
♻ ☆ More is Less: Inducing Sparsity via Overparameterization
In deep learning it is common to overparameterize neural networks, that is, to use more parameters than training samples. Quite surprisingly training the neural network via (stochastic) gradient descent leads to models that generalize very well, while classical statistics would suggest overfitting. In order to gain understanding of this implicit bias phenomenon we study the special case of sparse recovery (compressed sensing) which is of interest on its own. More precisely, in order to reconstruct a vector from underdetermined linear measurements, we introduce a corresponding overparameterized square loss functional, where the vector to be reconstructed is deeply factorized into several vectors. We show that, if there exists an exact solution, vanilla gradient flow for the overparameterized loss functional converges to a good approximation of the solution of minimal $\ell_1$-norm. The latter is well-known to promote sparse solutions. As a by-product, our results significantly improve the sample complexity for compressed sensing via gradient flow/descent on overparameterized models derived in previous works. The theory accurately predicts the recovery rate in numerical experiments. Our proof relies on analyzing a certain Bregman divergence of the flow. This bypasses the obstacles caused by non-convexity and should be of independent interest.
♻ ☆ The Power of Regularization in Solving Extensive-Form Games
In this paper, we investigate the power of {\it regularization}, a common technique in reinforcement learning and optimization, in solving extensive-form games (EFGs). We propose a series of new algorithms based on regularizing the payoff functions of the game, and establish a set of convergence results that strictly improve over the existing ones, with either weaker assumptions or stronger convergence guarantees. In particular, we first show that dilated optimistic mirror descent (DOMD), an efficient variant of OMD for solving EFGs, with adaptive regularization can achieve a fast $\tilde O(1/T)$ last-iterate convergence in terms of duality gap and distance to the set of Nash equilibrium (NE) without uniqueness assumption of the NE. Second, we show that regularized counterfactual regret minimization (\texttt{Reg-CFR}), with a variant of optimistic mirror descent algorithm as regret-minimizer, can achieve $O(1/T^{1/4})$ best-iterate, and $O(1/T^{3/4})$ average-iterate convergence rate for finding NE in EFGs. Finally, we show that \texttt{Reg-CFR} can achieve asymptotic last-iterate convergence, and optimal $O(1/T)$ average-iterate convergence rate, for finding the NE of perturbed EFGs, which is useful for finding approximate extensive-form perfect equilibria (EFPE). To the best of our knowledge, they constitute the first last-iterate convergence results for CFR-type algorithms, while matching the state-of-the-art average-iterate convergence rate in finding NE for non-perturbed EFGs. We also provide numerical results to corroborate the advantages of our algorithms.
♻ ☆ Continual Learning for Monolingual End-to-End Automatic Speech Recognition
Adapting Automatic Speech Recognition (ASR) models to new domains results in a deterioration of performance on the original domain(s), a phenomenon called Catastrophic Forgetting (CF). Even monolingual ASR models cannot be extended to new accents, dialects, topics, etc. without suffering from CF, making them unable to be continually enhanced without storing all past data. Fortunately, Continual Learning (CL) methods, which aim to enable continual adaptation while overcoming CF, can be used. In this paper, we implement an extensive number of CL methods for End-to-End ASR and test and compare their ability to extend a monolingual Hybrid CTC-Transformer model across four new tasks. We find that the best performing CL method closes the gap between the fine-tuned model (lower bound) and the model trained jointly on all tasks (upper bound) by more than 40%, while requiring access to only 0.6% of the original data.
comment: Published at EUSIPCO 2022. 5 pages, 1 figure
♻ ☆ Measuring Non-Probabilistic Uncertainty
There are two reasons why uncertainty about the future yield of investments may not be adequately described by Probability Theory. The first one is due to unique or nearly-unique events, that either never realized or occurred too seldom for probabilities to be reliable. The second one arises when when one fears that something may happen, that one is not even able to figure out, e.g., if one asks: "Climate change, financial crises, pandemic, war, what next?" In both cases, simple one-to-one causal mappings between available alternatives and possible consequences eventually melt down. However, such destructions reflect into the changing narratives of business executives, employees and other stakeholders in specific, identifiable and differential ways. In particular, texts such as consultants' reports or letters to shareholders can be analysed in order to detect the impact of both sorts of uncertainty onto the causal relations that normally guide decision-making. We propose structural measures of causal mappings as a means to measure non-probabilistic uncertainty, eventually suggesting that automated text analysis can greatly augment the possibilities offered by these techniques. Prospective applications may concern statistical institutes, stock market traders, as well as businesses wishing to compare their own vision to those prevailing in their industry.
comment: 22 pages, 15 figures
♻ ☆ Aerial View Localization with Reinforcement Learning: Towards Emulating Search-and-Rescue ICLR 2023
Climate-induced disasters are and will continue to be on the rise, and thus search-and-rescue (SAR) operations, where the task is to localize and assist one or several people who are missing, become increasingly relevant. In many cases the rough location may be known and a UAV can be deployed to explore a given, confined area to precisely localize the missing people. Due to time and battery constraints it is often critical that localization is performed as efficiently as possible. In this work we approach this type of problem by abstracting it as an aerial view goal localization task in a framework that emulates a SAR-like setup without requiring access to actual UAVs. In this framework, an agent operates on top of an aerial image (proxy for a search area) and is tasked with localizing a goal that is described in terms of visual cues. To further mimic the situation on an actual UAV, the agent is not able to observe the search area in its entirety, not even at low resolution, and thus it has to operate solely based on partial glimpses when navigating towards the goal. To tackle this task, we propose AiRLoc, a reinforcement learning (RL)-based model that decouples exploration (searching for distant goals) and exploitation (localizing nearby goals). Extensive evaluations show that AiRLoc outperforms heuristic search methods as well as alternative learnable approaches, and that it generalizes across datasets, e.g. to disaster-hit areas without seeing a single disaster scenario during training. We also conduct a proof-of-concept study which indicates that the learnable methods outperform humans on average. Code and models have been made publicly available at https://github.com/aleksispi/airloc.
comment: Accepted to ICLR 2023 Workshop on Machine Learning for Remote Sensing
♻ ☆ Optimal Algorithms for Latent Bandits with Cluster Structure AISTATS 2023
We consider the problem of latent bandits with cluster structure where there are multiple users, each with an associated multi-armed bandit problem. These users are grouped into \emph{latent} clusters such that the mean reward vectors of users within the same cluster are identical. At each round, a user, selected uniformly at random, pulls an arm and observes a corresponding noisy reward. The goal of the users is to maximize their cumulative rewards. This problem is central to practical recommendation systems and has received wide attention of late \cite{gentile2014online, maillard2014latent}. Now, if each user acts independently, then they would have to explore each arm independently and a regret of $\Omega(\sqrt{\mathsf{MNT}})$ is unavoidable, where $\mathsf{M}, \mathsf{N}$ are the number of arms and users, respectively. Instead, we propose LATTICE (Latent bAndiTs via maTrIx ComplEtion) which allows exploitation of the latent cluster structure to provide the minimax optimal regret of $\widetilde{O}(\sqrt{(\mathsf{M}+\mathsf{N})\mathsf{T}})$, when the number of clusters is $\widetilde{O}(1)$. This is the first algorithm to guarantee such strong regret bound. LATTICE is based on a careful exploitation of arm information within a cluster while simultaneously clustering users. Furthermore, it is computationally efficient and requires only $O(\log{\mathsf{T}})$ calls to an offline matrix completion oracle across all $\mathsf{T}$ rounds.
comment: 44 pages. Accepted to AISTATS 2023. Added Experiments
♻ ☆ Towards Good Practices in Evaluating Transfer Adversarial Attacks
Transfer adversarial attacks raise critical security concerns in real-world, black-box scenarios. However, the actual progress of this field is difficult to assess due to two common limitations in existing evaluations. First, different methods are often not systematically and fairly evaluated in a one-to-one comparison. Second, only transferability is evaluated but another key attack property, stealthiness, is largely overlooked. In this work, we design good practices to address these limitations, and we present the first comprehensive evaluation of transfer attacks, covering 23 representative attacks against 9 defenses on ImageNet. In particular, we propose to categorize existing attacks into five categories, which enables our systematic category-wise analyses. These analyses lead to new findings that even challenge existing knowledge and also help determine the optimal attack hyperparameters for our attack-wise comprehensive evaluation. We also pay particular attention to stealthiness, by adopting diverse imperceptibility metrics and looking into new, finer-grained characteristics. Overall, our new insights into transferability and stealthiness lead to actionable good practices for future evaluations.
comment: Our code and a list of categorized attacks are publicly available at https://github.com/ZhengyuZhao/TransferAttackEval
♻ ☆ Hair and Scalp Disease Detection using Machine Learning and Image Processing
Almost 80 million Americans suffer from hair loss due to aging, stress, medication, or genetic makeup. Hair and scalp-related diseases often go unnoticed in the beginning. Sometimes, a patient cannot differentiate between hair loss and regular hair fall. Diagnosing hair-related diseases is time-consuming as it requires professional dermatologists to perform visual and medical tests. Because of that, the overall diagnosis gets delayed, which worsens the severity of the illness. Due to the image-processing ability, neural network-based applications are used in various sectors, especially healthcare and health informatics, to predict deadly diseases like cancers and tumors. These applications assist clinicians and patients and provide an initial insight into early-stage symptoms. In this study, we used a deep learning approach that successfully predicts three main types of hair loss and scalp-related diseases: alopecia, psoriasis, and folliculitis. However, limited study in this area, unavailability of a proper dataset, and degree of variety among the images scattered over the internet made the task challenging. 150 images were obtained from various sources and then preprocessed by denoising, image equalization, enhancement, and data balancing, thereby minimizing the error rate. After feeding the processed data into the 2D convolutional neural network (CNN) model, we obtained overall training accuracy of 96.2%, with a validation accuracy of 91.1%. The precision and recall score of alopecia, psoriasis, and folliculitis are 0.895, 0.846, and 1.0, respectively. We also created a dataset of the scalp images for future prospective researchers.
♻ ☆ SE(3)-DiffusionFields: Learning smooth cost functions for joint grasp and motion optimization through diffusion
Multi-objective optimization problems are ubiquitous in robotics, e.g., the optimization of a robot manipulation task requires a joint consideration of grasp pose configurations, collisions and joint limits. While some demands can be easily hand-designed, e.g., the smoothness of a trajectory, several task-specific objectives need to be learned from data. This work introduces a method for learning data-driven SE(3) cost functions as diffusion models. Diffusion models can represent highly-expressive multimodal distributions and exhibit proper gradients over the entire space due to their score-matching training objective. Learning costs as diffusion models allows their seamless integration with other costs into a single differentiable objective function, enabling joint gradient-based motion optimization. In this work, we focus on learning SE(3) diffusion models for 6DoF grasping, giving rise to a novel framework for joint grasp and motion optimization without needing to decouple grasp selection from trajectory generation. We evaluate the representation power of our SE(3) diffusion models w.r.t. classical generative models, and we showcase the superior performance of our proposed optimization framework in a series of simulated and real-world robotic manipulation tasks against representative baselines.
comment: diffusion models, SE(3), grasping,
♻ ☆ On the Robustness of Dataset Inference
Machine learning (ML) models are costly to train as they can require a significant amount of data, computational resources and technical expertise. Thus, they constitute valuable intellectual property that needs protection from adversaries wanting to steal them. Ownership verification techniques allow the victims of model stealing attacks to demonstrate that a suspect model was in fact stolen from theirs. Although a number of ownership verification techniques based on watermarking or fingerprinting have been proposed, most of them fall short either in terms of security guarantees (well-equipped adversaries can evade verification) or computational cost. A fingerprinting technique introduced at ICLR '21, Dataset Inference (DI), has been shown to offer better robustness and efficiency than prior methods. The authors of DI provided a correctness proof for linear (suspect) models. However, in the same setting, we prove that DI suffers from high false positives (FPs) -- it can incorrectly identify an independent model trained with non-overlapping data from the same distribution as stolen. We further prove that DI also triggers FPs in realistic, non-linear suspect models. We then confirm empirically that DI leads to FPs, with high confidence. Second, we show that DI also suffers from false negatives (FNs) -- an adversary can fool DI by regularising a stolen model's decision boundaries using adversarial training, thereby leading to an FN. To this end, we demonstrate that DI fails to identify a model adversarially trained from a stolen dataset -- the setting where DI is the hardest to evade. Finally, we discuss the implications of our findings, the viability of fingerprinting-based ownership verification in general, and suggest directions for future work.
comment: 17 pages, 5 tables, 4 figures
♻ ☆ Efficient Recovery Learning using Model Predictive Meta-Reasoning ICRA
Operating under real world conditions is challenging due to the possibility of a wide range of failures induced by execution errors and state uncertainty. In relatively benign settings, such failures can be overcome by retrying or executing one of a small number of hand-engineered recovery strategies. By contrast, contact-rich sequential manipulation tasks, like opening doors and assembling furniture, are not amenable to exhaustive hand-engineering. To address this issue, we present a general approach for robustifying manipulation strategies in a sample-efficient manner. Our approach incrementally improves robustness by first discovering the failure modes of the current strategy via exploration in simulation and then learning additional recovery skills to handle these failures. To ensure efficient learning, we propose an online algorithm called Meta-Reasoning for Skill Learning (MetaReSkill) that monitors the progress of all recovery policies during training and allocates training resources to recoveries that are likely to improve the task performance the most. We use our approach to learn recovery skills for door-opening and evaluate them both in simulation and on a real robot with little fine-tuning. Compared to open-loop execution, our experiments show that even a limited amount of recovery learning improves task success substantially from 71% to 92.4% in simulation and from 75% to 90% on a real robot.
comment: To appear in the International Conference on Robotics and Automation (ICRA) 2023
♻ ☆ DIFFormer: Scalable (Graph) Transformers Induced by Energy Constrained Diffusion ICLR 2023
Real-world data generation often involves complex inter-dependencies among instances, violating the IID-data hypothesis of standard learning paradigms and posing a challenge for uncovering the geometric structures for learning desired instance representations. To this end, we introduce an energy constrained diffusion model which encodes a batch of instances from a dataset into evolutionary states that progressively incorporate other instances' information by their interactions. The diffusion process is constrained by descent criteria w.r.t.~a principled energy function that characterizes the global consistency of instance representations over latent structures. We provide rigorous theory that implies closed-form optimal estimates for the pairwise diffusion strength among arbitrary instance pairs, which gives rise to a new class of neural encoders, dubbed as DIFFormer (diffusion-based Transformers), with two instantiations: a simple version with linear complexity for prohibitive instance numbers, and an advanced version for learning complex structures. Experiments highlight the wide applicability of our model as a general-purpose encoder backbone with superior performance in various tasks, such as node classification on large graphs, semi-supervised image/text classification, and spatial-temporal dynamics prediction.
comment: Published at ICLR 2023 as a spotlight presentation, the implementation code is available at https://github.com/qitianwu/DIFFormer
♻ ☆ Decoupling Skill Learning from Robotic Control for Generalizable Object Manipulation ICRA
Recent works in robotic manipulation through reinforcement learning (RL) or imitation learning (IL) have shown potential for tackling a range of tasks e.g., opening a drawer or a cupboard. However, these techniques generalize poorly to unseen objects. We conjecture that this is due to the high-dimensional action space for joint control. In this paper, we take an alternative approach and separate the task of learning 'what to do' from 'how to do it' i.e., whole-body control. We pose the RL problem as one of determining the skill dynamics for a disembodied virtual manipulator interacting with articulated objects. The whole-body robotic kinematic control is optimized to execute the high-dimensional joint motion to reach the goals in the workspace. It does so by solving a quadratic programming (QP) model with robotic singularity and kinematic constraints. Our experiments on manipulating complex articulated objects show that the proposed approach is more generalizable to unseen objects with large intra-class variations, outperforming previous approaches. The evaluation results indicate that our approach generates more compliant robotic motion and outperforms the pure RL and IL baselines in task success rates. Additional information and videos are available at https://kl-research.github.io/decoupskill
comment: Accepted to IEEE International Conference on Robotics and Automation (ICRA) 2023
♻ ☆ Energy-based Out-of-Distribution Detection for Graph Neural Networks ICLR 2023
Learning on graphs, where instance nodes are inter-connected, has become one of the central problems for deep learning, as relational structures are pervasive and induce data inter-dependence which hinders trivial adaptation of existing approaches that assume inputs to be i.i.d.~sampled. However, current models mostly focus on improving testing performance of in-distribution data and largely ignore the potential risk w.r.t. out-of-distribution (OOD) testing samples that may cause negative outcome if the prediction is overconfident on them. In this paper, we investigate the under-explored problem, OOD detection on graph-structured data, and identify a provably effective OOD discriminator based on an energy function directly extracted from graph neural networks trained with standard classification loss. This paves a way for a simple, powerful and efficient OOD detection model for GNN-based learning on graphs, which we call GNNSafe. It also has nice theoretical properties that guarantee an overall distinguishable margin between the detection scores for in-distribution and OOD samples, which, more critically, can be further strengthened by a learning-free energy belief propagation scheme. For comprehensive evaluation, we introduce new benchmark settings that evaluate the model for detecting OOD data from both synthetic and real distribution shifts (cross-domain graph shifts and temporal graph shifts). The results show that GNNSafe achieves up to $17.0\%$ AUROC improvement over state-of-the-arts and it could serve as simple yet strong baselines in such an under-developed area.
comment: Published at ICLR 2023, the implementation code is available at https://github.com/qitianwu/GraphOOD-GNNSafe
♻ ☆ Dish-TS: A General Paradigm for Alleviating Distribution Shift in Time Series Forecasting AAAI 2023
The distribution shift in Time Series Forecasting (TSF), indicating series distribution changes over time, largely hinders the performance of TSF models. Existing works towards distribution shift in time series are mostly limited in the quantification of distribution and, more importantly, overlook the potential shift between lookback and horizon windows. To address above challenges, we systematically summarize the distribution shift in TSF into two categories. Regarding lookback windows as input-space and horizon windows as output-space, there exist (i) intra-space shift, that the distribution within the input-space keeps shifted over time, and (ii) inter-space shift, that the distribution is shifted between input-space and output-space. Then we introduce, Dish-TS, a general neural paradigm for alleviating distribution shift in TSF. Specifically, for better distribution estimation, we propose the coefficient net (CONET), which can be any neural architectures, to map input sequences into learnable distribution coefficients. To relieve intra-space and inter-space shift, we organize Dish-TS as a Dual-CONET framework to separately learn the distribution of input- and output-space, which naturally captures the distribution difference of two spaces. In addition, we introduce a more effective training strategy for intractable CONET learning. Finally, we conduct extensive experiments on several datasets coupled with different state-of-the-art forecasting models. Experimental results show Dish-TS consistently boosts them with a more than 20% average improvement. Code is available.
comment: Accepted by AAAI 2023
♻ ☆ Feature Affinity Assisted Knowledge Distillation and Quantization of Deep Neural Networks on Label-Free Data
In this paper, we propose a feature affinity (FA) assisted knowledge distillation (KD) method to improve quantization-aware training of deep neural networks (DNN). The FA loss on intermediate feature maps of DNNs plays the role of teaching middle steps of a solution to a student instead of only giving final answers in the conventional KD where the loss acts on the network logits at the output level. Combining logit loss and FA loss, we found that the quantized student network receives stronger supervision than from the labeled ground-truth data. The resulting FAQD is capable of compressing model on label-free data, which brings immediate practical benefits as pre-trained teacher models are readily available and unlabeled data are abundant. In contrast, data labeling is often laborious and expensive. Finally, we propose a fast feature affinity (FFA) loss that accurately approximates FA loss with a lower order of computational complexity, which helps speed up training for high resolution image input.
♻ ☆ Semantics-Native Communication with Contextual Reasoning
Spurred by a huge interest in the post-Shannon communication, it has recently been shown that leveraging semantics can significantly improve the communication effectiveness across many tasks. In this article, inspired by human communication, we propose a novel stochastic model of System 1 semantics-native communication (SNC) for generic tasks, where a speaker has an intention of referring to an entity, extracts the semantics, and communicates its symbolic representation to a target listener. To further reach its full potential, we additionally infuse contextual reasoning into SNC such that the speaker locally and iteratively self-communicates with a virtual agent built on the physical listener's unique way of coding its semantics, i.e., communication context. The resultant System 2 SNC allows the speaker to extract the most effective semantics for its listener. Leveraging the proposed stochastic model, we show that the reliability of System 2 SNC increases with the number of meaningful concepts, and derive the expected semantic representation (SR) bit length which quantifies the extracted effective semantics. It is also shown that System 2 SNC significantly reduces the SR length without compromising communication reliability.
comment: 18 pages, 16 figures, in IEEE Transactions on Cognitive Communications and Networking
♻ ☆ Indiscriminate Poisoning Attacks on Unsupervised Contrastive Learning ICLR 2023
Indiscriminate data poisoning attacks are quite effective against supervised learning. However, not much is known about their impact on unsupervised contrastive learning (CL). This paper is the first to consider indiscriminate poisoning attacks of contrastive learning. We propose Contrastive Poisoning (CP), the first effective such attack on CL. We empirically show that Contrastive Poisoning, not only drastically reduces the performance of CL algorithms, but also attacks supervised learning models, making it the most generalizable indiscriminate poisoning attack. We also show that CL algorithms with a momentum encoder are more robust to indiscriminate poisoning, and propose a new countermeasure based on matrix completion. Code is available at: https://github.com/kaiwenzha/contrastive-poisoning.
comment: ICLR 2023 Spotlight (notable top 25%). The first two authors contributed equally to this paper
♻ ☆ Learn2Agree: Fitting with Multiple Annotators without Objective Ground Truth ML4H
The annotation of domain experts is important for some medical applications where the objective ground truth is ambiguous to define, e.g., the rehabilitation for some chronic diseases, and the prescreening of some musculoskeletal abnormalities without further medical examinations. However, improper uses of the annotations may hinder developing reliable models. On one hand, forcing the use of a single ground truth generated from multiple annotations is less informative for the modeling. On the other hand, feeding the model with all the annotations without proper regularization is noisy given existing disagreements. For such issues, we propose a novel Learning to Agreement (Learn2Agree) framework to tackle the challenge of learning from multiple annotators without objective ground truth. The framework has two streams, with one stream fitting with the multiple annotators and the other stream learning agreement information between annotators. In particular, the agreement learning stream produces regularization information to the classifier stream, tuning its decision to be better in line with the agreement between annotators. The proposed method can be easily added to existing backbones, with experiments on two medical datasets showed better agreement levels with annotators.
comment: Accepted by the TML4H workshop at ICLR 2023
♻ ☆ Why (and When) does Local SGD Generalize Better than SGD? ICLR 2023
Local SGD is a communication-efficient variant of SGD for large-scale training, where multiple GPUs perform SGD independently and average the model parameters periodically. It has been recently observed that Local SGD can not only achieve the design goal of reducing the communication overhead but also lead to higher test accuracy than the corresponding SGD baseline (Lin et al., 2020b), though the training regimes for this to happen are still in debate (Ortiz et al., 2021). This paper aims to understand why (and when) Local SGD generalizes better based on Stochastic Differential Equation (SDE) approximation. The main contributions of this paper include (i) the derivation of an SDE that captures the long-term behavior of Local SGD in the small learning rate regime, showing how noise drives the iterate to drift and diffuse after it has reached close to the manifold of local minima, (ii) a comparison between the SDEs of Local SGD and SGD, showing that Local SGD induces a stronger drift term that can result in a stronger effect of regularization, e.g., a faster reduction of sharpness, and (iii) empirical evidence validating that having a small learning rate and long enough training time enables the generalization improvement over SGD but removing either of the two conditions leads to no improvement.
comment: Published as a conference paper at ICLR 2023
♻ ☆ A Survey on Federated Recommendation Systems
Federated learning has recently been applied to recommendation systems to protect user privacy. In federated learning settings, recommendation systems can train recommendation models only collecting the intermediate parameters instead of the real user data, which greatly enhances the user privacy. Beside, federated recommendation systems enable to collaborate with other data platforms to improve recommended model performance while meeting the regulation and privacy constraints. However, federated recommendation systems faces many new challenges such as privacy, security, heterogeneity and communication costs. While significant research has been conducted in these areas, gaps in the surveying literature still exist. In this survey, we-(1) summarize some common privacy mechanisms used in federated recommendation systems and discuss the advantages and limitations of each mechanism; (2) review some robust aggregation strategies and several novel attacks against security; (3) summarize some approaches to address heterogeneity and communication costs problems; (4)introduce some open source platforms that can be used to build federated recommendation systems; (5) present some prospective research directions in the future. This survey can guide researchers and practitioners understand the research progress in these areas.
♻ ☆ Low-Rank Tensor Completion With Generalized CP Decomposition and Nonnegative Integer Tensor Completion
Tensor completion is important to many areas such as computer vision, data analysis, and signal processing. Previously, a category of methods known as low-rank tensor completion has been proposed and developed, involving the enforcement of low-rank structures on completed tensors. While such methods have been constantly improved, none considered exploiting the numerical properties of tensor elements. This work attempts to construct a new methodological framework called GCDTC (Generalized CP Decomposition Tensor Completion) based on numerical properties to achieve higher accuracy in tensor completion. In this newly introduced framework, a generalized form of the CP Decomposition is applied to low-rank tensor completion. This paper also proposes an algorithm known as SPTC (Smooth Poisson Tensor Completion) for nonnegative integer tensor completion as an application of the GCDTC framework. Through experimentation with real-life data, it is verified that this method could produce results superior in completion accuracy to current state-of-the-art methodologies.
comment: 10 pages, 4 figures, and 1 table
♻ ☆ Causal Confusion and Reward Misidentification in Preference-Based Reward Learning ICLR 2023
Learning policies via preference-based reward learning is an increasingly popular method for customizing agent behavior, but has been shown anecdotally to be prone to spurious correlations and reward hacking behaviors. While much prior work focuses on causal confusion in reinforcement learning and behavioral cloning, we focus on a systematic study of causal confusion and reward misidentification when learning from preferences. In particular, we perform a series of sensitivity and ablation analyses on several benchmark domains where rewards learned from preferences achieve minimal test error but fail to generalize to out-of-distribution states -- resulting in poor policy performance when optimized. We find that the presence of non-causal distractor features, noise in the stated preferences, and partial state observability can all exacerbate reward misidentification. We also identify a set of methods with which to interpret misidentified learned rewards. In general, we observe that optimizing misidentified rewards drives the policy off the reward's training distribution, resulting in high predicted (learned) rewards but low true rewards. These findings illuminate the susceptibility of preference learning to reward misidentification and causal confusion -- failure to consider even one of many factors can result in unexpected, undesirable behavior.
comment: In the proceedings of the Eleventh International Conference on Learning Representations (ICLR 2023). $\href{https://iclr.cc/virtual/2023/poster/10822}{\text{URL}}$
♻ ☆ Group Fairness in Non-monotone Submodular Maximization
Maximizing a submodular function has a wide range of applications in machine learning and data mining. One such application is data summarization whose goal is to select a small set of representative and diverse data items from a large dataset. However, data items might have sensitive attributes such as race or gender, in this setting, it is important to design \emph{fairness-aware} algorithms to mitigate potential algorithmic bias that may cause over- or under- representation of particular groups. Motivated by that, we propose and study the classic non-monotone submodular maximization problem subject to novel group fairness constraints. Our goal is to select a set of items that maximizes a non-monotone submodular function, while ensuring that the number of selected items from each group is proportionate to its size, to the extent specified by the decision maker. We develop the first constant-factor approximation algorithms for this problem. We also extend the basic model to incorporate an additional global size constraint on the total number of selected items.
comment: This article has been accepted for publication in the Journal on Combinatorial Optimization
♻ ☆ On the Risks of Stealing the Decoding Algorithms of Language Models
A key component of generating text from modern language models (LM) is the selection and tuning of decoding algorithms. These algorithms determine how to generate text from the internal probability distribution generated by the LM. The process of choosing a decoding algorithm and tuning its hyperparameters takes significant time, manual effort, and computation, and it also requires extensive human evaluation. Therefore, the identity and hyperparameters of such decoding algorithms are considered to be extremely valuable to their owners. In this work, we show, for the first time, that an adversary with typical API access to an LM can steal the type and hyperparameters of its decoding algorithms at very low monetary costs. Our attack is effective against popular LMs used in text generation APIs, including GPT-2 and GPT-3. We demonstrate the feasibility of stealing such information with only a few dollars, e.g., $\$0.8$, $\$1$, $\$4$, and $\$40$ for the four versions of GPT-3.
♻ ☆ Near Optimal Memory-Regret Tradeoff for Online Learning
In the experts problem, on each of $T$ days, an agent needs to follow the advice of one of $n$ ``experts''. After each day, the loss associated with each expert's advice is revealed. A fundamental result in learning theory says that the agent can achieve vanishing regret, i.e. their cumulative loss is within $o(T)$ of the cumulative loss of the best-in-hindsight expert. Can the agent perform well without sufficient space to remember all the experts? We extend a nascent line of research on this question in two directions: $\bullet$ We give a new algorithm against the oblivious adversary, improving over the memory-regret tradeoff obtained by [PZ23], and nearly matching the lower bound of [SWXZ22]. $\bullet$ We also consider an adaptive adversary who can observe past experts chosen by the agent. In this setting we give both a new algorithm and a novel lower bound, proving that roughly $\sqrt{n}$ memory is both necessary and sufficient for obtaining $o(T)$ regret.
♻ ☆ Asynchronous Hybrid Reinforcement Learning for Latency and Reliability Optimization in the Metaverse over Wireless Communications
Technology advancements in wireless communications and high-performance Extended Reality (XR) have empowered the developments of the Metaverse. The demand for the Metaverse applications and hence, real-time digital twinning of real-world scenes is increasing. Nevertheless, the replication of 2D physical world images into 3D virtual objects is computationally intensive and requires computation offloading. The disparity in transmitted object dimension (2D as opposed to 3D) leads to asymmetric data sizes in uplink (UL) and downlink (DL). To ensure the reliability and low latency of the system, we consider an asynchronous joint UL-DL scenario where in the UL stage, the smaller data size of the physical world images captured by multiple extended reality users (XUs) will be uploaded to the Metaverse Console (MC) to be construed and rendered. In the DL stage, the larger-size 3D virtual objects need to be transmitted back to the XUs. We design a novel multi-agent reinforcement learning algorithm structure, namely Asynchronous Actors Hybrid Critic (AAHC), to optimize the decisions pertaining to computation offloading and channel assignment in the UL stage and optimize the DL transmission power in the DL stage. Extensive experiments demonstrate that compared to proposed baselines, AAHC obtains better solutions with satisfactory training time.
comment: This paper appears in IEEE Journal on Selected Areas in Communications (JSAC), 2023
♻ ☆ CorruptEncoder: Data Poisoning based Backdoor Attacks to Contrastive Learning
Contrastive learning (CL) pre-trains general-purpose encoders using an unlabeled pre-training dataset, which consists of images or image-text pairs. CL is vulnerable to data poisoning based backdoor attacks (DPBAs), in which an attacker injects poisoned inputs into the pre-training dataset so the encoder is backdoored. However, existing DPBAs achieve limited effectiveness. In this work, we propose new DPBAs called CorruptEncoder to CL. CorruptEncoder uses a theory-guided method to create optimal poisoned inputs to maximize attack effectiveness. Our experiments show that CorruptEncoder substantially outperforms existing DPBAs. In particular, CorruptEncoder is the first DPBA that achieves more than 90% attack success rates with only a few (3) reference images and a small poisoning ratio (0.5%). Moreover, we also propose a defense, called localized cropping, to defend against DPBAs. Our results show that our defense can reduce the effectiveness of DPBAs, though it slightly sacrifices the utility of the encoder.
♻ ☆ Benchmarking AutoML algorithms on a collection of synthetic classification problems
Automated machine learning (AutoML) algorithms have grown in popularity due to their high performance and flexibility to adapt to different problems and data sets. With the increasing number of AutoML algorithms, deciding which would best suit a given problem becomes increasingly more work. Therefore, it is essential to use complex and challenging benchmarks which would be able to differentiate the AutoML algorithms from each other. This paper compares the performance of four different AutoML algorithms: Tree-based Pipeline Optimization Tool (TPOT), Auto-Sklearn, Auto-Sklearn 2, and H2O AutoML. We use the Diverse and Generative ML benchmark (DIGEN), a diverse set of synthetic datasets derived from generative functions designed to highlight the strengths and weaknesses of the performance of common machine learning algorithms. We confirm that AutoML can identify pipelines that perform well on all included datasets. Most AutoML algorithms performed similarly; however, there were some differences depending on the specific dataset and metric used.
♻ ☆ Part-Based Models Improve Adversarial Robustness ICLR 2023
We show that combining human prior knowledge with end-to-end learning can improve the robustness of deep neural networks by introducing a part-based model for object classification. We believe that the richer form of annotation helps guide neural networks to learn more robust features without requiring more samples or larger models. Our model combines a part segmentation model with a tiny classifier and is trained end-to-end to simultaneously segment objects into parts and then classify the segmented object. Empirically, our part-based models achieve both higher accuracy and higher adversarial robustness than a ResNet-50 baseline on all three datasets. For instance, the clean accuracy of our part models is up to 15 percentage points higher than the baseline's, given the same level of robustness. Our experiments indicate that these models also reduce texture bias and yield better robustness against common corruptions and spurious correlations. The code is publicly available at https://github.com/chawins/adv-part-model.
comment: Published in ICLR 2023 (poster). Code can be found at https://github.com/chawins/adv-part-model
Predictive Inference with Feature Conformal Prediction ICLR 2023
Conformal prediction is a distribution-free technique for establishing valid prediction intervals. Although conventionally people conduct conformal prediction in the output space, this is not the only possibility. In this paper, we propose feature conformal prediction, which extends the scope of conformal prediction to semantic feature spaces by leveraging the inductive bias of deep representation learning. From a theoretical perspective, we demonstrate that feature conformal prediction provably outperforms regular conformal prediction under mild assumptions. Our approach could be combined with not only vanilla conformal prediction, but also other adaptive conformal prediction methods. Apart from experiments on existing predictive inference benchmarks, we also demonstrate the state-of-the-art performance of the proposed methods on large-scale tasks such as ImageNet classification and Cityscapes image segmentation.
comment: Published as a conference paper at ICLR 2023
♻ ☆ TDSTF: Transformer-based Diffusion probabilistic model for Sparse Time series Forecasting
\noindent \textbf{Background and objective:} In the intensive care unit (ICU), vital sign monitoring is critical, and an accurate predictive system is required. This study will create a novel model to forecast Heart Rate (HR), Systolic Blood Pressure (SBP), and Diastolic Blood Pressure (DBP) in ICU. These vital signs are crucial for prompt interventions for patients. We extracted $24,886$ ICU stays from the MIMIC-III database, which contains data from over $46$ thousand patients, to train and test the model. \noindent \textbf{Methods:} The model proposed in this study, areansformerin intensive careabilistic Model for Sparse Time Series Forecasting (TDSTF), uses a deep learning technique called the Transformer. The TDSTF model showed state-of-the-art performance in predicting vital signs in the ICU, outperforming other models' ability to predict distributions of vital signs and being more computationally efficient. The code is available at https://github.com/PingChang818/TDSTF. \noindent \textbf{Results:} The results of the study showed that TDSTF achieved a Normalized Average Continuous Ranked Probability Score (NACRPS) of $0.4438$ and a Mean Squared Error (MSE) of $0.4168$, an improvement of $18.9\%$ and $34.3\%$ over the best baseline model, respectively. \noindent \textbf{Conclusion:} In conclusion, TDSTF is an effective and efficient solution for forecasting vital signs in the ICU, and it shows a significant improvement compared to other models in the field. \noindent \textbf{Keywords: deep learning, time series forecasting, sparse data, vital signs, ICU}
♻ ☆ Faster Adaptive Federated Learning
Federated learning has attracted increasing attention with the emergence of distributed data. While extensive federated learning algorithms have been proposed for the non-convex distributed problem, federated learning in practice still faces numerous challenges, such as the large training iterations to converge since the sizes of models and datasets keep increasing, and the lack of adaptivity by SGD-based model updates. Meanwhile, the study of adaptive methods in federated learning is scarce and existing works either lack a complete theoretical convergence guarantee or have slow sample complexity. In this paper, we propose an efficient adaptive algorithm (i.e., FAFED) based on the momentum-based variance-reduced technique in cross-silo FL. We first explore how to design the adaptive algorithm in the FL setting. By providing a counter-example, we prove that a simple combination of FL and adaptive methods could lead to divergence. More importantly, we provide a convergence analysis for our method and prove that our algorithm is the first adaptive FL algorithm to reach the best-known samples $O(\epsilon^{-3})$ and $O(\epsilon^{-2})$ communication rounds to find an $\epsilon$-stationary point without large batches. The experimental results on the language modeling task and image classification task with heterogeneous data demonstrate the efficiency of our algorithms.
♻ ☆ Validation of a Hospital Digital Twin with Machine Learning
Recently there has been a surge of interest in developing Digital Twins of process flows in healthcare to better understand bottlenecks and areas of improvement. A key challenge is in the validation process. We describe a work in progress for a digital twin using an agent based simulation model for determining bed turnaround time for patients in hospitals. We employ a strategy using machine learning for validating the model and implementing sensitivity analysis.
♻ ☆ Dataset of Random Relaxations for Crystal Structure Search of Li-Si System
Crystal structure search is a long-standing challenge in materials design. We present a dataset of more than 100,000 structural relaxations of potential battery anode materials from randomized structures using density functional theory calculations. We illustrate the usage of the dataset by training graph neural networks to predict structural relaxations from randomly generated structures. Our models directly predict stresses in addition to forces, which allows them to accurately simulate relaxations of both ionic positions and lattice vectors. We show that models trained on the molecular dynamics simulations fail to simulate relaxations from random structures, while training on our data leads to up to two orders of magnitude decrease in error for the same task. Our model is able to find an experimentally verified structure of a stoichiometry held out from training. We find that randomly perturbing atomic positions during training improves both the accuracy and out of domain generalization of the models.
♻ ☆ Convergence Rates of Stochastic Zeroth-order Gradient Descent for Ł ojasiewicz Functions
We prove convergence rates of Stochastic Zeroth-order Gradient Descent (SZGD) algorithms for Lojasiewicz functions. The SZGD algorithm iterates as \begin{align*} \mathbf{x}_{t+1} = \mathbf{x}_t - \eta_t \widehat{\nabla} f (\mathbf{x}_t), \qquad t = 0,1,2,3,\cdots , \end{align*} where $f$ is the objective function that satisfies the \L ojasiewicz inequality with \L ojasiewicz exponent $\theta$, $\eta_t$ is the step size (learning rate), and $ \widehat{\nabla} f (\mathbf{x}_t) $ is the approximate gradient estimated using zeroth-order information only. Our results show that $ \{ f (\mathbf{x}_t) - f (\mathbf{x}_\infty) \}_{t \in \mathbb{N} } $ can converge faster than $ \{ \| \mathbf{x}_t - \mathbf{x}_\infty \| \}_{t \in \mathbb{N} }$, regardless of whether the objective $f$ is smooth or nonsmooth.
comment: V3: more than major revision. Y. Feng is added to the author list. V4: length cut and some typo corrections
♻ ☆ Backdoor Detection and Mitigation in Competitive Reinforcement Learning
While real-world applications of reinforcement learning are becoming popular, the security and robustness of RL systems are worthy of more attention and exploration. In particular, recent works have revealed that, in a multi-agent RL environment, backdoor trigger actions can be injected into a victim agent (a.k.a. Trojan agent), which can result in a catastrophic failure as soon as it sees the backdoor trigger action. To ensure the security of RL agents against malicious backdoors, in this work, we propose the problem of Backdoor Detection in a multi-agent competitive reinforcement learning system, with the objective of detecting Trojan agents as well as the corresponding potential trigger actions, and further trying to mitigate their Trojan behavior. In order to solve this problem, we propose PolicyCleanse that is based on the property that the activated Trojan agents accumulated rewards degrade noticeably after several timesteps. Along with PolicyCleanse, we also design a machine unlearning-based approach that can effectively mitigate the detected backdoor. Extensive experiments demonstrate that the proposed methods can accurately detect Trojan agents, and outperform existing backdoor mitigation baseline approaches by at least 3% in winning rate across various types of agents and environments.
♻ ☆ Pishgu: Universal Path Prediction Network Architecture for Real-time Cyber-physical Edge Systems
Path prediction is an essential task for many real-world Cyber-Physical Systems (CPS) applications, from autonomous driving and traffic monitoring/management to pedestrian/worker safety. These real-world CPS applications need a robust, lightweight path prediction that can provide a universal network architecture for multiple subjects (e.g., pedestrians and vehicles) from different perspectives. However, most existing algorithms are tailor-made for a unique subject with a specific camera perspective and scenario. This article presents Pishgu, a universal lightweight network architecture, as a robust and holistic solution for path prediction. Pishgu's architecture can adapt to multiple path prediction domains with different subjects (vehicles, pedestrians), perspectives (bird's-eye, high-angle), and scenes (sidewalk, highway). Our proposed architecture captures the inter-dependencies within the subjects in each frame by taking advantage of Graph Isomorphism Networks and the attention module. We separately train and evaluate the efficacy of our architecture on three different CPS domains across multiple perspectives (vehicle bird's-eye view, pedestrian bird's-eye view, and human high-angle view). Pishgu outperforms state-of-the-art solutions in the vehicle bird's-eye view domain by 42% and 61% and pedestrian high-angle view domain by 23% and 22% in terms of ADE and FDE, respectively. Additionally, we analyze the domain-specific details for various datasets to understand their effect on path prediction and model interpretation. Finally, we report the latency and throughput for all three domains on multiple embedded platforms showcasing the robustness and adaptability of Pishgu for real-world integration into CPS applications.
♻ ☆ Efficient recurrent architectures through activity sparsity and sparse back-propagation through time ICLR 2023
Recurrent neural networks (RNNs) are well suited for solving sequence tasks in resource-constrained systems due to their expressivity and low computational requirements. However, there is still a need to bridge the gap between what RNNs are capable of in terms of efficiency and performance and real-world application requirements. The memory and computational requirements arising from propagating the activations of all the neurons at every time step to every connected neuron, together with the sequential dependence of activations, contribute to the inefficiency of training and using RNNs. We propose a solution inspired by biological neuron dynamics that makes the communication between RNN units sparse and discrete. This makes the backward pass with backpropagation through time (BPTT) computationally sparse and efficient as well. We base our model on the gated recurrent unit (GRU), extending it with units that emit discrete events for communication triggered by a threshold so that no information is communicated to other units in the absence of events. We show theoretically that the communication between units, and hence the computation required for both the forward and backward passes, scales with the number of events in the network. Our model achieves efficiency without compromising task performance, demonstrating competitive performance compared to state-of-the-art recurrent network models in real-world tasks, including language modeling. The dynamic activity sparsity mechanism also makes our model well suited for novel energy-efficient neuromorphic hardware. Code is available at https://github.com/KhaleelKhan/EvNN/.
comment: Published as notable-top-25% paper in ICLR 2023
♻ ☆ Guaranteed Conformance of Neurosymbolic Models to Natural Constraints
Deep neural networks have emerged as the workhorse for a large section of robotics and control applications, especially as models for dynamical systems. Such data-driven models are in turn used for designing and verifying autonomous systems. This is particularly useful in modeling medical systems where data can be leveraged to individualize treatment. In safety-critical applications, it is important that the data-driven model is conformant to established knowledge from the natural sciences. Such knowledge is often available or can often be distilled into a (possibly black-box) model $M$. For instance, the unicycle model (which encodes Newton's laws) for an F1 racing car. In this light, we consider the following problem - given a model $M$ and state transition dataset, we wish to best approximate the system model while being bounded distance away from $M$. We propose a method to guarantee this conformance. Our first step is to distill the dataset into few representative samples called memories, using the idea of a growing neural gas. Next, using these memories we partition the state space into disjoint subsets and compute bounds that should be respected by the neural network, when the input is drawn from a particular subset. This serves as a symbolic wrapper for guaranteed conformance. We argue theoretically that this only leads to bounded increase in approximation error; which can be controlled by increasing the number of memories. We experimentally show that on three case studies (Car Model, Drones, and Artificial Pancreas), our constrained neurosymbolic models conform to specified $M$ models (each encoding various constraints) with order-of-magnitude improvements compared to the augmented Lagrangian and vanilla training methods. Our code can be found at https://github.com/kaustubhsridhar/Constrained_Models
♻ ☆ Low Dimensional Invariant Embeddings for Universal Geometric Learning
This paper studies separating invariants: mappings on $D$ dimensional domains which are invariant to an appropriate group action, and which separate orbits. The motivation for this study comes from the usefulness of separating invariants in proving universality of equivariant neural network architectures. We observe that in several cases the cardinality of separating invariants proposed in the machine learning literature is much larger than the dimension $D$. As a result, the theoretical universal constructions based on these separating invariants is unrealistically large. Our goal in this paper is to resolve this issue. We show that when a continuous family of semi-algebraic separating invariants is available, separation can be obtained by randomly selecting 2D+1 of these invariants. We apply this methodology to obtain an efficient scheme for computing separating invariants for several classical group actions which have been studied in the invariant learning literature. Examples include matrix multiplication actions on point clouds by permutations, rotations, and various other linear groups. Often the requirement of invariant separation is relaxed and only generic separation is required. In this case, we show that only D+1 invariants are required. More importantly, generic invariants are often significantly easier to compute, as we illustrate by discussing generic and full separation for weighted graphs. Finally we outline an approach for proving that separating invariants can be constructed also when the random parameters have finite precision.
♻ ☆ A Typology for Exploring the Mitigation of Shortcut Behavior
As machine learning models become increasingly larger, trained weakly supervised on large, possibly uncurated data sets, it becomes increasingly important to establish mechanisms for inspecting, interacting, and revising models to mitigate learning shortcuts and guarantee their learned knowledge is aligned with human knowledge. The recently proposed XIL framework was developed for this purpose, and several such methods have been introduced, each with individual motivations and methodological details. In this work, we provide a unification of various XIL methods into a single typology by establishing a common set of basic modules. In doing so, we pave the way for a principled comparison of existing, but, importantly, also future XIL approaches. In addition, we discuss existing and introduce novel measures and benchmarks for evaluating the overall abilities of a XIL method. Given this extensive toolbox, including our typology, measures, and benchmarks, we finally compare several recent XIL methods methodologically and quantitatively. In our evaluations, all methods prove to revise a model successfully. However, we found remarkable differences in individual benchmark tasks, revealing valuable application-relevant aspects for integrating these benchmarks in developing future methods.
♻ ☆ Learning POD of Complex Dynamics Using Heavy-ball Neural ODEs
Proper orthogonal decomposition (POD) allows reduced-order modeling of complex dynamical systems at a substantial level, while maintaining a high degree of accuracy in modeling the underlying dynamical systems. Advances in machine learning algorithms enable learning POD-based dynamics from data and making accurate and fast predictions of dynamical systems. In this paper, we leverage the recently proposed heavy-ball neural ODEs (HBNODEs) [Xia et al. NeurIPS, 2021] for learning data-driven reduced-order models (ROMs) in the POD context, in particular, for learning dynamics of time-varying coefficients generated by the POD analysis on training snapshots generated from solving full order models. HBNODE enjoys several practical advantages for learning POD-based ROMs with theoretical guarantees, including 1) HBNODE can learn long-term dependencies effectively from sequential observations and 2) HBNODE is computationally efficient in both training and testing. We compare HBNODE with other popular ROMs on several complex dynamical systems, including the von K\'{a}rm\'{a}n Street flow, the Kurganov-Petrova-Popov equation, and the one-dimensional Euler equations for fluids modeling.
comment: 31 pages, 20 figures
♻ ☆ Linked Data Science Powered by Knowledge Graphs
In recent years, we have witnessed a growing interest in data science not only from academia but particularly from companies investing in data science platforms to analyze large amounts of data. In this process, a myriad of data science artifacts, such as datasets and pipeline scripts, are created. Yet, there has so far been no systematic attempt to holistically exploit the collected knowledge and experiences that are implicitly contained in the specification of these pipelines, e.g., compatible datasets, cleansing steps, ML algorithms, parameters, etc. Instead, data scientists still spend a considerable amount of their time trying to recover relevant information and experiences from colleagues, trial and error, lengthy exploration, etc. In this paper, we, therefore, propose a scalable system (KGLiDS) that employs machine learning to extract the semantics of data science pipelines and captures them in a knowledge graph, which can then be exploited to assist data scientists in various ways. This abstraction is the key to enabling Linked Data Science since it allows us to share the essence of pipelines between platforms, companies, and institutions without revealing critical internal information and instead focusing on the semantics of what is being processed and how. Our comprehensive evaluation uses thousands of datasets and more than thirteen thousand pipeline scripts extracted from data discovery benchmarks and the Kaggle portal and shows that KGLiDS significantly outperforms state-of-the-art systems on related tasks, such as dataset recommendation and pipeline classification.
comment: 11 pages, 6 figures
♻ ☆ Adaptive Activation-based Structured Pruning
Pruning is a promising approach to compress complex deep learning models in order to deploy them on resource-constrained edge devices. However, many existing pruning solutions are based on unstructured pruning, which yields models that cannot efficiently run on commodity hardware and require users to manually explore and tune the pruning process, which is time-consuming and often leads to sub-optimal results. To address these limitations, this paper presents an adaptive, activation-based, structured pruning approach to automatically and efficiently generate small, accurate, and hardware-efficient models that meet user requirements. First, it proposes iterative structured pruning using activation-based attention feature maps to effectively identify and prune unimportant filters. Then, it proposes adaptive pruning policies for automatically meeting the pruning objectives of accuracy-critical, memory-constrained, and latency-sensitive tasks. A comprehensive evaluation shows that the proposed method can substantially outperform the state-of-the-art structured pruning works on CIFAR-10 and ImageNet datasets. For example, on ResNet-56 with CIFAR-10, without any accuracy drop, our method achieves the largest parameter reduction (79.11%), outperforming the related works by 22.81% to 66.07%, and the largest FLOPs reduction (70.13%), outperforming the related works by 14.13% to 26.53%.
♻ ☆ Tensor Denoising via Amplification and Stable Rank Methods
Tensors in the form of multilinear arrays are ubiquitous in data science applications. Captured real-world data, including video, hyperspectral images, and discretized physical systems, naturally occur as tensors and often come with attendant noise. Under the additive noise model and with the assumption that the underlying clean tensor has low rank, many denoising methods have been created that utilize tensor decomposition to effect denoising through low rank tensor approximation. However, all such decomposition methods require estimating the tensor rank, or related measures such as the tensor spectral and nuclear norms, all of which are NP-hard problems. In this work we leverage our previously developed framework of $\textit{tensor amplification}$, which provides good approximations of the spectral and nuclear tensor norms, to denoising synthetic tensors of various sizes, ranks, and noise levels, along with real-world tensors derived from physiological signals. We also introduce two new notions of tensor rank -- $\textit{stable slice rank}$ and $\textit{stable }$$X$$\textit{-rank}$ -- and new denoising methods based on their estimation. The experimental results show that in the low rank context, tensor-based amplification provides comparable denoising performance in high signal-to-noise ratio (SNR) settings and superior performance in noisy (i.e., low SNR) settings, while the stable $X$-rank method achieves superior denoising performance on the physiological signal data.
Multimedia
☆ MMCosine: Multi-Modal Cosine Loss Towards Balanced Audio-Visual Fine-Grained Learning
Audio-visual learning helps to comprehensively understand the world by fusing practical information from multiple modalities. However, recent studies show that the imbalanced optimization of uni-modal encoders in a joint-learning model is a bottleneck to enhancing the model's performance. We further find that the up-to-date imbalance-mitigating methods fail on some audio-visual fine-grained tasks, which have a higher demand for distinguishable feature distribution. Fueled by the success of cosine loss that builds hyperspherical feature spaces and achieves lower intra-class angular variability, this paper proposes Multi-Modal Cosine loss, MMCosine. It performs a modality-wise $L_2$ normalization to features and weights towards balanced and better multi-modal fine-grained learning. We demonstrate that our method can alleviate the imbalanced optimization from the perspective of weight norm and fully exploit the discriminability of the cosine metric. Extensive experiments prove the effectiveness of our method and the versatility with advanced multi-modal fusion strategies and up-to-date imbalance-mitigating methods.
☆ Improving Few-Shot Learning for Talking Face System with TTS Data Augmentation ICASSP 2023
Audio-driven talking face has attracted broad interest from academia and industry recently. However, data acquisition and labeling in audio-driven talking face are labor-intensive and costly. The lack of data resource results in poor synthesis effect. To alleviate this issue, we propose to use TTS (Text-To-Speech) for data augmentation to improve few-shot ability of the talking face system. The misalignment problem brought by the TTS audio is solved with the introduction of soft-DTW, which is first adopted in the talking face task. Moreover, features extracted by HuBERT are explored to utilize underlying information of audio, and found to be superior over other features. The proposed method achieves 17%, 14%, 38% dominance on MSE score, DTW score and user study preference repectively over the baseline model, which shows the effectiveness of improving few-shot learning for talking face system with TTS augmentation.
comment: 4 pages. Accepted by ICASSP 2023
☆ Towards Robust Image-in-Audio Deep Steganography
The field of steganography has experienced a surge of interest due to the recent advancements in AI-powered techniques, particularly in the context of multimodal setups that enable the concealment of signals within signals of a different nature. The primary objectives of all steganographic methods are to achieve perceptual transparency, robustness, and large embedding capacity - which often present conflicting goals that classical methods have struggled to reconcile. This paper extends and enhances an existing image-in-audio deep steganography method by focusing on improving its robustness. The proposed enhancements include modifications to the loss function, utilization of the Short-Time Fourier Transform (STFT), introduction of redundancy in the encoding process for error correction, and buffering of additional information in the pixel subconvolution operation. The results demonstrate that our approach outperforms the existing method in terms of robustness and perceptual transparency.
comment: 8 pages, 5 figures, 2 tables
♻ ☆ BIRD-PCC: Bi-directional Range Image-based Deep LiDAR Point Cloud Compression ICASSP 2023
The large amount of data collected by LiDAR sensors brings the issue of LiDAR point cloud compression (PCC). Previous works on LiDAR PCC have used range image representations and followed the predictive coding paradigm to create a basic prototype of a coding framework. However, their prediction methods give an inaccurate result due to the negligence of invalid pixels in range images and the omission of future frames in the time step. Moreover, their handcrafted design of residual coding methods could not fully exploit spatial redundancy. To remedy this, we propose a coding framework BIRD-PCC. Our prediction module is aware of the coordinates of invalid pixels in range images and takes a bidirectional scheme. Also, we introduce a deep-learned residual coding module that can further exploit spatial redundancy within a residual frame. Experiments conducted on SemanticKITTI and KITTI-360 datasets show that BIRD-PCC outperforms other methods in most bitrate conditions and generalizes well to unseen environments.
comment: Accepted to ICASSP 2023
Computation and Language
☆ Comprehensive Event Representations using Event Knowledge Graphs and Natural Language Processing
Recent work has utilised knowledge-aware approaches to natural language understanding, question answering, recommendation systems, and other tasks. These approaches rely on well-constructed and large-scale knowledge graphs that can be useful for many downstream applications and empower knowledge-aware models with commonsense reasoning. Such knowledge graphs are constructed through knowledge acquisition tasks such as relation extraction and knowledge graph completion. This work seeks to utilise and build on the growing body of work that uses findings from the field of natural language processing (NLP) to extract knowledge from text and build knowledge graphs. The focus of this research project is on how we can use transformer-based approaches to extract and contextualise event information, matching it to existing ontologies, to build a comprehensive knowledge of graph-based event representations. Specifically, sub-event extraction is used as a way of creating sub-event-aware event representations. These event representations are then further enriched through fine-grained location extraction and contextualised through the alignment of historically relevant quotes.
comment: This is the author's version of the work. It is posted here for your personal use. Not for redistribution. The definitive Version of Record was published in Companion Proceedings of the Web Conference 2022
☆ On the Risks of Stealing the Decoding Algorithms of Language Models
A key component of generating text from modern language models (LM) is the selection and tuning of decoding algorithms. These algorithms determine how to generate text from the internal probability distribution generated by the LM. The process of choosing a decoding algorithm and tuning its hyperparameters takes significant time, manual effort, and computation, and it also requires extensive human evaluation. Therefore, the identity and hyperparameters of such decoding algorithms are considered to be extremely valuable to their owners. In this work, we show, for the first time, that an adversary with typical API access to an LM can steal the type and hyperparameters of its decoding algorithms at very low monetary costs. Our attack is effective against popular LMs used in text generation APIs, including GPT-2 and GPT-3. We demonstrate the feasibility of stealing such information with only a few dollars, e.g., $\$0.8$, $\$1$, $\$4$, and $\$40$ for the four versions of GPT-3.
☆ Extending the Pre-Training of BLOOM for Improved Support of Traditional Chinese: Models, Methods and Results
In this paper we present the multilingual language model BLOOM-zh that features enhanced support for Traditional Chinese. BLOOM-zh has its origins in the open-source BLOOM models presented by BigScience in 2022. Starting from released models, we extended the pre-training of BLOOM by additional 7.4 billion tokens in Traditional Chinese and English covering a variety of domains such as news articles, books, encyclopedias, educational materials as well as spoken language. In order to show the properties of BLOOM-zh, both existing and newly created benchmark scenarios are used for evaluating the performance. BLOOM-zh outperforms its predecessor on most Traditional Chinese benchmarks while maintaining its English capability. We release all our models to the research community.
☆ Self-contained Beta-with-Spikes Approximation for Inference Under a Wright-Fisher Model
We construct a reliable estimation of evolutionary parameters within the Wright-Fisher model, which describes changes in allele frequencies due to selection and genetic drift, from time-series data. Such data exists for biological populations, for example via artificial evolution experiments, and for the cultural evolution of behavior, such as linguistic corpora that document historical usage of different words with similar meanings. Our method of analysis builds on a Beta-with-Spikes approximation to the distribution of allele frequencies predicted by the Wright-Fisher model. We introduce a self-contained scheme for estimating the parameters in the approximation, and demonstrate its robustness with synthetic data, especially in the strong-selection and near-extinction regimes where previous approaches fail. We further apply to allele frequency data for baker's yeast (Saccharomyces cerevisiae), finding a significant signal of selection in cases where independent evidence supports such a conclusion. We further demonstrate the possibility of detecting time-points at which evolutionary parameters change in the context of a historical spelling reform in the Spanish language.
☆ Cost-Effective Hyperparameter Optimization for Large Language Model Generation Inference
Large Language Models (LLMs) like GPT-3 have sparked significant interest in their generative capabilities, leading to the development of various commercial applications. The high cost of using the models drives application builders to maximize the value of generation under a limited inference budget. This paper presents a study of optimizing inference hyperparameters like the number of responses, temperature and max tokens, which significantly affects the utility/cost of text generation. We design a framework named EcoOptiGen which leverages economical hyperparameter optimization and cost-based pruning. Experiments with the latest GPT-3.5 models on a variety of tasks verify its effectiveness. EcoOptiGen is implemented in the FLAML library: https://github.com/microsoft/FLAML, and we provide one example of using it at: https://microsoft.github.io/FLAML/docs/Examples/Integrate%20-%20OpenAI.
☆ Extrapolative Controlled Sequence Generation via Iterative Refinement
We study the problem of extrapolative controlled generation, i.e., generating sequences with attribute values beyond the range seen in training. This task is of significant importance in automated design, especially drug discovery, where the goal is to design novel proteins that are \textit{better} (e.g., more stable) than existing sequences. Thus, by definition, the target sequences and their attribute values are out of the training distribution, posing challenges to existing methods that aim to directly generate the target sequence. Instead, in this work, we propose Iterative Controlled Extrapolation (ICE) which iteratively makes local edits to a sequence to enable extrapolation. We train the model on synthetically generated sequence pairs that demonstrate small improvement in the attribute value. Results on one natural language task (sentiment analysis) and two protein engineering tasks (ACE2 stability and AAV fitness) show that ICE considerably outperforms state-of-the-art approaches despite its simplicity. Our code and models are available at: https://github.com/vishakhpk/iter-extrapolation.
comment: Preprint
☆ Models of symbol emergence in communication: a conceptual review and a guide for avoiding local minima
Computational simulations are a popular method for testing hypotheses about the emergence of communication. This kind of research is performed in a variety of traditions including language evolution, developmental psychology, cognitive science, machine learning, robotics, etc. The motivations for the models are different, but the operationalizations and methods used are often similar. We identify the assumptions and explanatory targets of several most representative models and summarise the known results. We claim that some of the assumptions -- such as portraying meaning in terms of mapping, focusing on the descriptive function of communication, modelling signals with amodal tokens -- may hinder the success of modelling. Relaxing these assumptions and foregrounding the interactions of embodied and situated agents allows one to systematise the multiplicity of pressures under which symbolic systems evolve. In line with this perspective, we sketch the road towards modelling the emergence of meaningful symbolic communication, where symbols are simultaneously grounded in action and perception and form an abstract system.
☆ Student's t-Distribution: On Measuring the Inter-Rater Reliability When the Observations are Scarce
In natural language processing (NLP) we always rely on human judgement as the golden quality evaluation method. However, there has been an ongoing debate on how to better evaluate inter-rater reliability (IRR) levels for certain evaluation tasks, such as translation quality evaluation (TQE), especially when the data samples (observations) are very scarce. In this work, we first introduce the study on how to estimate the confidence interval for the measurement value when only one data (evaluation) point is available. Then, this leads to our example with two human-generated observational scores, for which, we introduce ``Student's \textit{t}-Distribution'' method and explain how to use it to measure the IRR score using only these two data points, as well as the confidence intervals (CIs) of the quality evaluation. We give quantitative analysis on how the evaluation confidence can be greatly improved by introducing more observations, even if only one extra observation. We encourage researchers to report their IRR scores in all possible means, e.g. using Student's \textit{t}-Distribution method whenever possible; thus making the NLP evaluation more meaningful, transparent, and trustworthy. This \textit{t}-Distribution method can be also used outside of NLP fields to measure IRR level for trustworthy evaluation of experimental investigations, whenever the observational data is scarce. Keywords: Inter-Rater Reliability (IRR); Scarce Observations; Confidence Intervals (CIs); Natural Language Processing (NLP); Translation Quality Evaluation (TQE); Student's \textit{t}-Distribution
comment: 10 pages
☆ MenuCraft: Interactive Menu System Design with Large Language Models
Menu system design is a challenging task involving many design options and various human factors. For example, one crucial factor that designers need to consider is the semantic and systematic relation of menu commands. However, capturing these relations can be challenging due to limited available resources. With the advancement of neural language models, large language models can utilize their vast pre-existing knowledge in designing and refining menu systems. In this paper, we propose MenuCraft, an AI-assisted designer for menu design that enables collaboration between the designer and a dialogue system to design menus. MenuCraft offers an interactive language-based menu design tool that simplifies the menu design process and enables easy customization of design options. MenuCraft supports a variety of interactions through dialog that allows performing few-shot learning.
☆ Query-Utterance Attention with Joint modeling for Query-Focused Meeting Summarization
Query-focused meeting summarization (QFMS) aims to generate summaries from meeting transcripts in response to a given query. Previous works typically concatenate the query with meeting transcripts and implicitly model the query relevance only at the token level with attention mechanism. However, due to the dilution of key query-relevant information caused by long meeting transcripts, the original transformer-based model is insufficient to highlight the key parts related to the query. In this paper, we propose a query-aware framework with joint modeling token and utterance based on Query-Utterance Attention. It calculates the utterance-level relevance to the query with a dense retrieval module. Then both token-level query relevance and utterance-level query relevance are combined and incorporated into the generation process with attention mechanism explicitly. We show that the query relevance of different granularities contributes to generating a summary more related to the query. Experimental results on the QMSum dataset show that the proposed model achieves new state-of-the-art performance.
comment: icassp 2023
☆ NASTyLinker: NIL-Aware Scalable Transformer-based Entity Linker ESWC'23
Entity Linking (EL) is the task of detecting mentions of entities in text and disambiguating them to a reference knowledge base. Most prevalent EL approaches assume that the reference knowledge base is complete. In practice, however, it is necessary to deal with the case of linking to an entity that is not contained in the knowledge base (NIL entity). Recent works have shown that, instead of focusing only on affinities between mentions and entities, considering inter-mention affinities can be used to represent NIL entities by producing clusters of mentions. At the same time, inter-mention affinities can help to substantially improve linking performance for known entities. With NASTyLinker, we introduce an EL approach that is aware of NIL-entities and produces corresponding mention clusters while maintaining high linking performance for known entities. The approach clusters mentions and entities based on dense representations from Transformers and resolves conflicts (if more than one entity is assigned to a cluster) by computing transitive mention-entity affinities. We show the effectiveness and scalability of NASTyLinker on NILK, a dataset that is explicitly constructed to evaluate EL with respect to NIL-entities. Further, we apply the presented approach to an actual EL task, namely to knowledge graph population by linking entities in Wikipedia listings, and provide an analysis of the outcome.
comment: Preprint of a paper in the research track of the 20th Extended Semantic Web Conference (ESWC'23)
☆ Automatically Auditing Large Language Models via Discrete Optimization
Auditing large language models for unexpected behaviors is critical to preempt catastrophic deployments, yet remains challenging. In this work, we cast auditing as an optimization problem, where we automatically search for input-output pairs that match a desired target behavior. For example, we might aim to find a non-toxic input that starts with "Barack Obama" that a model maps to a toxic output. This optimization problem is difficult to solve as the set of feasible points is sparse, the space is discrete, and the language models we audit are non-linear and high-dimensional. To combat these challenges, we introduce a discrete optimization algorithm, ARCA, that jointly and efficiently optimizes over inputs and outputs. Our approach automatically uncovers derogatory completions about celebrities (e.g. "Barack Obama is a legalized unborn" -> "child murderer"), produces French inputs that complete to English outputs, and finds inputs that generate a specific name. Our work offers a promising new tool to uncover models' failure-modes before deployment.
☆ Sample Efficient Multimodal Semantic Augmentation for Incremental Summarization
In this work, we develop a prompting approach for incremental summarization of task videos. We develop a sample-efficient few-shot approach for extracting semantic concepts as an intermediate step. We leverage an existing model for extracting the concepts from the images and extend it to videos and introduce a clustering and querying approach for sample efficiency, motivated by the recent advances in perceiver-based architectures. Our work provides further evidence that an approach with richer input context with relevant entities and actions from the videos and using these as prompts could enhance the summaries generated by the model. We show the results on a relevant dataset and discuss possible directions for the work.
☆ Does Synthetic Data Generation of LLMs Help Clinical Text Mining?
Recent advancements in large language models (LLMs) have led to the development of highly potent models like OpenAI's ChatGPT. These models have exhibited exceptional performance in a variety of tasks, such as question answering, essay composition, and code generation. However, their effectiveness in the healthcare sector remains uncertain. In this study, we seek to investigate the potential of ChatGPT to aid in clinical text mining by examining its ability to extract structured information from unstructured healthcare texts, with a focus on biological named entity recognition and relation extraction. However, our preliminary results indicate that employing ChatGPT directly for these tasks resulted in poor performance and raised privacy concerns associated with uploading patients' information to the ChatGPT API. To overcome these limitations, we propose a new training paradigm that involves generating a vast quantity of high-quality synthetic data with labels utilizing ChatGPT and fine-tuning a local model for the downstream task. Our method has resulted in significant improvements in the performance of downstream tasks, improving the F1-score from 23.37% to 63.99% for the named entity recognition task and from 75.86% to 83.59% for the relation extraction task. Furthermore, generating data using ChatGPT can significantly reduce the time and effort required for data collection and labeling, as well as mitigate data privacy concerns. In summary, the proposed framework presents a promising solution to enhance the applicability of LLM models to clinical text mining.
comment: 10 pages, 8 tables, 4 figures
☆ Lexical Complexity Prediction: An Overview
The occurrence of unknown words in texts significantly hinders reading comprehension. To improve accessibility for specific target populations, computational modelling has been applied to identify complex words in texts and substitute them for simpler alternatives. In this paper, we present an overview of computational approaches to lexical complexity prediction focusing on the work carried out on English data. We survey relevant approaches to this problem which include traditional machine learning classifiers (e.g. SVMs, logistic regression) and deep neural networks as well as a variety of features, such as those inspired by literature in psycholinguistics as well as word frequency, word length, and many others. Furthermore, we introduce readers to past competitions and available datasets created on this topic. Finally, we include brief sections on applications of lexical complexity prediction, such as readability and text simplification, together with related studies on languages other than English.
☆ The Casual Conversations v2 Dataset
This paper introduces a new large consent-driven dataset aimed at assisting in the evaluation of algorithmic bias and robustness of computer vision and audio speech models in regards to 11 attributes that are self-provided or labeled by trained annotators. The dataset includes 26,467 videos of 5,567 unique paid participants, with an average of almost 5 videos per person, recorded in Brazil, India, Indonesia, Mexico, Vietnam, Philippines, and the USA, representing diverse demographic characteristics. The participants agreed for their data to be used in assessing fairness of AI models and provided self-reported age, gender, language/dialect, disability status, physical adornments, physical attributes and geo-location information, while trained annotators labeled apparent skin tone using the Fitzpatrick Skin Type and Monk Skin Tone scales, and voice timbre. Annotators also labeled for different recording setups and per-second activity annotations.
☆ disco: a toolkit for Distributional Control of Generative Models
Pre-trained language models and other generative models have revolutionized NLP and beyond. However, these models tend to reproduce undesirable biases present in their training data. Also, they may overlook patterns that are important but challenging to capture. To address these limitations, researchers have introduced distributional control techniques. These techniques, not limited to language, allow controlling the prevalence (i.e., expectations) of any features of interest in the model's outputs. Despite their potential, the widespread adoption of these techniques has been hindered by the difficulty in adapting complex, disconnected code. Here, we present disco, an open-source Python library that brings these techniques to the broader public.
☆ Automatic Detection of Industry Sectors in Legal Articles Using Machine Learning Approaches
The ability to automatically identify industry sector coverage in articles on legal developments, or any kind of news articles for that matter, can bring plentiful of benefits both to the readers and the content creators themselves. By having articles tagged based on industry coverage, readers from all around the world would be able to get to legal news that are specific to their region and professional industry. Simultaneously, writers would benefit from understanding which industries potentially lack coverage or which industries readers are currently mostly interested in and thus, they would focus their writing efforts towards more inclusive and relevant legal news coverage. In this paper, a Machine Learning-powered industry analysis approach which combined Natural Language Processing (NLP) with Statistical and Machine Learning (ML) techniques was investigated. A dataset consisting of over 1,700 annotated legal articles was created for the identification of six industry sectors. Text and legal based features were extracted from the text. Both traditional ML methods (e.g. gradient boosting machine algorithms, and decision-tree based algorithms) and deep neural network (e.g. transformer models) were applied for performance comparison of predictive models. The system achieved promising results with area under the receiver operating characteristic curve scores above 0.90 and F-scores above 0.81 with respect to the six industry sectors. The experimental results show that the suggested automated industry analysis which employs ML techniques allows the processing of large collections of text data in an easy, efficient, and scalable way. Traditional ML methods perform better than deep neural networks when only a small and domain-specific training data is available for the study.
comment: 26 pages, 5 figures, 3 tables. Paper was presented at 'Classification and Data Science in the Digital Age', 17th conference of the International Federation of Classification Societies (IFCS2022), Porto, Portugal, https://ifcs2022.fep.up.pt/
♻ ☆ Lila: A Unified Benchmark for Mathematical Reasoning EMNLP 2022
Mathematical reasoning skills are essential for general-purpose intelligent systems to perform tasks from grocery shopping to climate modeling. Towards evaluating and improving AI systems in this domain, we propose LILA, a unified mathematical reasoning benchmark consisting of 23 diverse tasks along four dimensions: (i) mathematical abilities e.g., arithmetic, calculus (ii) language format e.g., question-answering, fill-in-the-blanks (iii) language diversity e.g., no language, simple language (iv) external knowledge e.g., commonsense, physics. We construct our benchmark by extending 20 datasets benchmark by collecting task instructions and solutions in the form of Python programs, thereby obtaining explainable solutions in addition to the correct answer. We additionally introduce two evaluation datasets to measure out-of-distribution performance and robustness to language perturbation. Finally, we introduce BHASKARA, a general-purpose mathematical reasoning model trained on LILA. Importantly, we find that multi-tasking leads to significant improvements (average relative improvement of 21.83% F1 score vs. single-task models), while the best performing model only obtains 60.40%, indicating the room for improvement in general mathematical reasoning and understanding.
comment: EMNLP 2022
♻ ☆ Weakly Supervised Concept Map Generation through Task-Guided Graph Translation
Recent years have witnessed the rapid development of concept map generation techniques due to their advantages in providing well-structured summarization of knowledge from free texts. Traditional unsupervised methods do not generate task-oriented concept maps, whereas deep generative models require large amounts of training data. In this work, we present GT-D2G (Graph Translation-based Document To Graph), an automatic concept map generation framework that leverages generalized NLP pipelines to derive semantic-rich initial graphs, and translates them into more concise structures under the weak supervision of downstream task labels. The concept maps generated by GT-D2G can provide interpretable summarization of structured knowledge for the input texts, which are demonstrated through human evaluation and case studies on three real-world corpora. Further experiments on the downstream task of document classification show that GT-D2G beats other concept map generation methods. Moreover, we specifically validate the labeling efficiency of GT-D2G in the label-efficient learning setting and the flexibility of generated graph sizes in controlled hyper-parameter studies.
comment: Accepted by IEEE TKDE. All code and data available at https://github.com/lujiaying/GT-doc2graph
♻ ☆ Domain Adaptation of Transformer-Based Models using Unlabeled Data for Relevance and Polarity Classification of German Customer Feedback
Understanding customer feedback is becoming a necessity for companies to identify problems and improve their products and services. Text classification and sentiment analysis can play a major role in analyzing this data by using a variety of machine and deep learning approaches. In this work, different transformer-based models are utilized to explore how efficient these models are when working with a German customer feedback dataset. In addition, these pre-trained models are further analyzed to determine if adapting them to a specific domain using unlabeled data can yield better results than off-the-shelf pre-trained models. To evaluate the models, two downstream tasks from the GermEval 2017 are considered. The experimental results show that transformer-based models can reach significant improvements compared to a fastText baseline and outperform the published scores and previous models. For the subtask Relevance Classification, the best models achieve a micro-averaged $F1$-Score of 96.1 % on the first test set and 95.9 % on the second one, and a score of 85.1 % and 85.3 % for the subtask Polarity Classification.
comment: Complete
♻ ☆ Towards Inferential Reproducibility of Machine Learning Research ICLR 2023
Reliability of machine learning evaluation -- the consistency of observed evaluation scores across replicated model training runs -- is affected by several sources of nondeterminism which can be regarded as measurement noise. Current tendencies to remove noise in order to enforce reproducibility of research results neglect inherent nondeterminism at the implementation level and disregard crucial interaction effects between algorithmic noise factors and data properties. This limits the scope of conclusions that can be drawn from such experiments. Instead of removing noise, we propose to incorporate several sources of variance, including their interaction with data properties, into an analysis of significance and reliability of machine learning evaluation, with the aim to draw inferences beyond particular instances of trained models. We show how to use linear mixed effects models (LMEMs) to analyze performance evaluation scores, and to conduct statistical inference with a generalized likelihood ratio test (GLRT). This allows us to incorporate arbitrary sources of noise like meta-parameter variations into statistical significance testing, and to assess performance differences conditional on data properties. Furthermore, a variance component analysis (VCA) enables the analysis of the contribution of noise sources to overall variance and the computation of a reliability coefficient by the ratio of substantial to total variance.
comment: Published at ICLR 2023 (see https://openreview.net/pdf?id=li4GQCQWkv)
♻ ☆ Grounding Language with Visual Affordances over Unstructured Data ICRA
Recent works have shown that Large Language Models (LLMs) can be applied to ground natural language to a wide variety of robot skills. However, in practice, learning multi-task, language-conditioned robotic skills typically requires large-scale data collection and frequent human intervention to reset the environment or help correcting the current policies. In this work, we propose a novel approach to efficiently learn general-purpose language-conditioned robot skills from unstructured, offline and reset-free data in the real world by exploiting a self-supervised visuo-lingual affordance model, which requires annotating as little as 1% of the total data with language. We evaluate our method in extensive experiments both in simulated and real-world robotic tasks, achieving state-of-the-art performance on the challenging CALVIN benchmark and learning over 25 distinct visuomotor manipulation tasks with a single policy in the real world. We find that when paired with LLMs to break down abstract natural language instructions into subgoals via few-shot prompting, our method is capable of completing long-horizon, multi-tier tasks in the real world, while requiring an order of magnitude less data than previous approaches. Code and videos are available at http://hulc2.cs.uni-freiburg.de
comment: Accepted at the 2023 IEEE International Conference on Robotics and Automation (ICRA). Project website: http://hulc2.cs.uni-freiburg.de
♻ ☆ Visual Language Maps for Robot Navigation ICRA
Grounding language to the visual observations of a navigating agent can be performed using off-the-shelf visual-language models pretrained on Internet-scale data (e.g., image captions). While this is useful for matching images to natural language descriptions of object goals, it remains disjoint from the process of mapping the environment, so that it lacks the spatial precision of classic geometric maps. To address this problem, we propose VLMaps, a spatial map representation that directly fuses pretrained visual-language features with a 3D reconstruction of the physical world. VLMaps can be autonomously built from video feed on robots using standard exploration approaches and enables natural language indexing of the map without additional labeled data. Specifically, when combined with large language models (LLMs), VLMaps can be used to (i) translate natural language commands into a sequence of open-vocabulary navigation goals (which, beyond prior work, can be spatial by construction, e.g., "in between the sofa and TV" or "three meters to the right of the chair") directly localized in the map, and (ii) can be shared among multiple robots with different embodiments to generate new obstacle maps on-the-fly (by using a list of obstacle categories). Extensive experiments carried out in simulated and real world environments show that VLMaps enable navigation according to more complex language instructions than existing methods. Videos are available at https://vlmaps.github.io.
comment: Accepted at the 2023 IEEE International Conference on Robotics and Automation (ICRA). Project page: https://vlmaps.github.io
♻ ☆ ChatGPT: Beginning of an End of Manual Linguistic Data Annotation? Use Case of Automatic Genre Identification
ChatGPT has shown strong capabilities in natural language generation tasks, which naturally leads researchers to explore where its abilities end. In this paper, we examine whether ChatGPT can be used for zero-shot text classification, more specifically, automatic genre identification. We compare ChatGPT with a multilingual XLM-RoBERTa language model that was fine-tuned on datasets, manually annotated with genres. The models are compared on test sets in two languages: English and Slovenian. Results show that ChatGPT outperforms the fine-tuned model when applied to the dataset which was not seen before by either of the models. Even when applied on Slovenian language as an under-resourced language, ChatGPT's performance is no worse than when applied to English. However, if the model is fully prompted in Slovenian, the performance drops significantly, showing the current limitations of ChatGPT usage on smaller languages. The presented results lead us to questioning whether this is the beginning of an end of laborious manual annotation campaigns even for smaller languages, such as Slovenian.
♻ ☆ Describe me an Aucklet: Generating Grounded Perceptual Category Descriptions
Human language users can generate descriptions of perceptual concepts beyond instance-level representations and also use such descriptions to learn provisional class-level representations. However, the ability of computational models to learn and operate with class representations is under-investigated in the language-and-vision field. In this paper, we train separate neural networks to generate and interpret class-level descriptions. We then use the zero-shot classification performance of the interpretation model as a measure of communicative success and class-level conceptual grounding. We investigate the performance of prototype- and exemplar-based neural representations grounded category description. Finally, we show that communicative success reveals performance issues in the generation model that are not captured by traditional intrinsic NLG evaluation metrics, and argue that these issues can be traced to a failure to properly ground language in vision at the class level. We observe that the interpretation model performs better with descriptions that are low in diversity on the class level, possibly indicating a strong reliance on frequently occurring features.
♻ ☆ Larger language models do in-context learning differently
We study how in-context learning (ICL) in language models is affected by semantic priors versus input-label mappings. We investigate two setups-ICL with flipped labels and ICL with semantically-unrelated labels-across various model families (GPT-3, InstructGPT, Codex, PaLM, and Flan-PaLM). First, experiments on ICL with flipped labels show that overriding semantic priors is an emergent ability of model scale. While small language models ignore flipped labels presented in-context and thus rely primarily on semantic priors from pretraining, large models can override semantic priors when presented with in-context exemplars that contradict priors, despite the stronger semantic priors that larger models may hold. We next study semantically-unrelated label ICL (SUL-ICL), in which labels are semantically unrelated to their inputs (e.g., foo/bar instead of negative/positive), thereby forcing language models to learn the input-label mappings shown in in-context exemplars in order to perform the task. The ability to do SUL-ICL also emerges primarily with scale, and large-enough language models can even perform linear classification in a SUL-ICL setting. Finally, we evaluate instruction-tuned models and find that instruction tuning strengthens both the use of semantic priors and the capacity to learn input-label mappings, but more of the former.
♻ ☆ VoP: Text-Video Co-operative Prompt Tuning for Cross-Modal Retrieval CVPR 2023
Many recent studies leverage the pre-trained CLIP for text-video cross-modal retrieval by tuning the backbone with additional heavy modules, which not only brings huge computational burdens with much more parameters, but also leads to the knowledge forgetting from upstream models.In this work, we propose the VoP: Text-Video Co-operative Prompt Tuning for efficient tuning on the text-video retrieval task. The proposed VoP is an end-to-end framework with both video & text prompts introducing, which can be regarded as a powerful baseline with only 0.1% trainable parameters. Further, based on the spatio-temporal characteristics of videos, we develop three novel video prompt mechanisms to improve the performance with different scales of trainable parameters. The basic idea of the VoP enhancement is to model the frame position, frame context, and layer function with specific trainable prompts, respectively. Extensive experiments show that compared to full fine-tuning, the enhanced VoP achieves a 1.4% average R@1 gain across five text-video retrieval benchmarks with 6x less parameter overhead. The code will be available at https://github.com/bighuang624/VoP.
comment: Accepted by CVPR 2023
♻ ☆ SumREN: Summarizing Reported Speech about Events in News AAAI 2023
A primary objective of news articles is to establish the factual record for an event, frequently achieved by conveying both the details of the specified event (i.e., the 5 Ws; Who, What, Where, When and Why regarding the event) and how people reacted to it (i.e., reported statements). However, existing work on news summarization almost exclusively focuses on the event details. In this work, we propose the novel task of summarizing the reactions of different speakers, as expressed by their reported statements, to a given event. To this end, we create a new multi-document summarization benchmark, SUMREN, comprising 745 summaries of reported statements from various public figures obtained from 633 news articles discussing 132 events. We propose an automatic silver training data generation approach for our task, which helps smaller models like BART achieve GPT-3 level performance on this task. Finally, we introduce a pipeline-based framework for summarizing reported speech, which we empirically show to generate summaries that are more abstractive and factual than baseline query-focused summarization approaches.
comment: Accepted at AAAI 2023
♻ ☆ Universal resources for quantum computing
Unravelling the source of quantum computing power has been a major goal in the field of quantum information science. In recent years, the quantum resource theory (QRT) has been established to characterize various quantum resources, yet their roles in quantum computing tasks still require investigation. The so-called universal quantum computing model (UQCM), e.g., the circuit model, has been the main framework to guide the design of quantum algorithms, creation of real quantum computers etc. In this work, we combine the study of UQCM together with QRT. We find on one hand, using QRT can provide a resource-theoretic characterization of a UQCM, the relation among models and inspire new ones, and on the other hand, using UQCM offers a framework to apply resources, study relation among resources and classify them. We develop the theory of universal resources in the setting of UQCM, and find a rich spectrum of UQCMs and the corresponding universal resources. Depending on a hierarchical structure of resource theories, we find models can be classified into families. In this work, we study three natural families of UQCMs in details: the amplitude family, the quasi-probability family, and the Hamiltonian family. They include some well known models, like the measurement-based model and adiabatic model, and also inspire new models such as the contextual model we introduce. Each family contains at least a triplet of models, and such a succinct structure of families of UQCMs offers a unifying picture to investigate resources and design models. It also provides a rigorous framework to resolve puzzles, such as the role of entanglement vs. interference, and unravel resource-theoretic features of quantum algorithms.
♻ ☆ Mask-guided BERT for Few Shot Text Classification
Transformer-based language models have achieved significant success in various domains. However, the data-intensive nature of the transformer architecture requires much labeled data, which is challenging in low-resource scenarios (i.e., few-shot learning (FSL)). The main challenge of FSL is the difficulty of training robust models on small amounts of samples, which frequently leads to overfitting. Here we present Mask-BERT, a simple and modular framework to help BERT-based architectures tackle FSL. The proposed approach fundamentally differs from existing FSL strategies such as prompt tuning and meta-learning. The core idea is to selectively apply masks on text inputs and filter out irrelevant information, which guides the model to focus on discriminative tokens that influence prediction results. In addition, to make the text representations from different categories more separable and the text representations from the same category more compact, we introduce a contrastive learning loss function. Experimental results on public-domain benchmark datasets demonstrate the effectiveness of Mask-BERT.
♻ ☆ Check Your Facts and Try Again: Improving Large Language Models with External Knowledge and Automated Feedback
Large language models (LLMs), such as ChatGPT, are able to generate human-like, fluent responses for many downstream tasks, e.g., task-oriented dialog and question answering. However, applying LLMs to real-world, mission-critical applications remains challenging mainly due to their tendency to generate hallucinations and their inability to use external knowledge. This paper proposes a LLM-Augmenter system, which augments a black-box LLM with a set of plug-and-play modules. Our system makes the LLM generate responses grounded in external knowledge, e.g., stored in task-specific databases. It also iteratively revises LLM prompts to improve model responses using feedback generated by utility functions, e.g., the factuality score of a LLM-generated response. The effectiveness of LLM-Augmenter is empirically validated on two types of scenarios, task-oriented dialog and open-domain question answering. LLM-Augmenter significantly reduces ChatGPT's hallucinations without sacrificing the fluency and informativeness of its responses. We make the source code and models publicly available.
comment: 15 pages
♻ ☆ kogito: A Commonsense Knowledge Inference Toolkit EACL 2023
In this paper, we present kogito, an open-source tool for generating commonsense inferences about situations described in text. kogito provides an intuitive and extensible interface to interact with natural language generation models that can be used for hypothesizing commonsense knowledge inference from a textual input. In particular, kogito offers several features for targeted, multi-granularity knowledge generation. These include a standardized API for training and evaluating knowledge models, and generating and filtering inferences from them. We also include helper functions for converting natural language texts into a format ingestible by knowledge models - intermediate pipeline stages such as knowledge head extraction from text, heuristic and model-based knowledge head-relation matching, and an ability to define and use custom knowledge relations. We make the code for kogito available at https://github.com/epfl-nlp/kogito along with thorough documentation at https://kogito.readthedocs.io.
comment: EACL 2023 Camera ready, 9 pages
♻ ☆ A data science and machine learning approach to continuous analysis of Shakespeare's plays
The availability of quantitative methods that can analyze text has provided new ways of examining literature in a manner that was not available in the pre-information era. Here we apply comprehensive machine learning analysis to the work of William Shakespeare. The analysis shows clear change in style of writing over time, with the most significant changes in the sentence length, frequency of adjectives and adverbs, and the sentiments expressed in the text. Applying machine learning to make a stylometric prediction of the year of the play shows a Pearson correlation of 0.71 between the actual and predicted year, indicating that Shakespeare's writing style as reflected by the quantitative measurements changed over time. Additionally, it shows that the stylometrics of some of the plays is more similar to plays written either before or after the year they were written. For instance, Romeo and Juliet is dated 1596, but is more similar in stylometrics to plays written by Shakespeare after 1600. The source code for the analysis is available for free download.
comment: Journal of Data Mining and Digital Humanities, accepted
♻ ☆ Fillers in Spoken Language Understanding: Computational and Psycholinguistic Perspectives
Disfluencies (i.e. interruptions in the regular flow of speech), are ubiquitous to spoken discourse. Fillers ("uh", "um") are disfluencies that occur the most frequently compared to other kinds of disfluencies. Yet, to the best of our knowledge, there isn't a resource that brings together the research perspectives influencing Spoken Language Understanding (SLU) on these speech events. This aim of this article is to survey a breadth of perspectives in a holistic way; i.e. from considering underlying (psycho)linguistic theory, to their annotation and consideration in Automatic Speech Recognition (ASR) and SLU systems, to lastly, their study from a generation standpoint. This article aims to present the perspectives in an approachable way to the SLU and Conversational AI community, and discuss moving forward, what we believe are the trends and challenges in each area.
comment: To appear in TAL Journal
♻ ☆ Automatic Context Pattern Generation for Entity Set Expansion
Entity Set Expansion (ESE) is a valuable task that aims to find entities of the target semantic class described by given seed entities. Various Natural Language Processing (NLP) and Information Retrieval (IR) downstream applications have benefited from ESE due to its ability to discover knowledge. Although existing corpus-based ESE methods have achieved great progress, they still rely on corpora with high-quality entity information annotated, because most of them need to obtain the context patterns through the position of the entity in a sentence. Therefore, the quality of the given corpora and their entity annotation has become the bottleneck that limits the performance of such methods. To overcome this dilemma and make the ESE models free from the dependence on entity annotation, our work aims to explore a new ESE paradigm, namely corpus-independent ESE. Specifically, we devise a context pattern generation module that utilizes autoregressive language models (e.g., GPT-2) to automatically generate high-quality context patterns for entities. In addition, we propose the GAPA, a novel ESE framework that leverages the aforementioned GenerAted PAtterns to expand target entities. Extensive experiments and detailed analyses on three widely used datasets demonstrate the effectiveness of our method. All the codes of our experiments are available at https://github.com/geekjuruo/GAPA.
comment: This work has been submitted to the IEEE for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessible
Computer Vision and Pattern Recognition
☆ X-Avatar: Expressive Human Avatars
We present X-Avatar, a novel avatar model that captures the full expressiveness of digital humans to bring about life-like experiences in telepresence, AR/VR and beyond. Our method models bodies, hands, facial expressions and appearance in a holistic fashion and can be learned from either full 3D scans or RGB-D data. To achieve this, we propose a part-aware learned forward skinning module that can be driven by the parameter space of SMPL-X, allowing for expressive animation of X-Avatars. To efficiently learn the neural shape and deformation fields, we propose novel part-aware sampling and initialization strategies. This leads to higher fidelity results, especially for smaller body parts while maintaining efficient training despite increased number of articulated bones. To capture the appearance of the avatar with high-frequency details, we extend the geometry and deformation fields with a texture network that is conditioned on pose, facial expression, geometry and the normals of the deformed surface. We show experimentally that our method outperforms strong baselines in both data domains both quantitatively and qualitatively on the animation task. To facilitate future research on expressive avatars we contribute a new dataset, called X-Humans, containing 233 sequences of high-quality textured scans from 20 participants, totalling 35,500 data frames.
comment: Project page: https://skype-line.github.io/projects/X-Avatar/
☆ Open-Vocabulary Panoptic Segmentation with Text-to-Image Diffusion Models CVPR 2022
We present ODISE: Open-vocabulary DIffusion-based panoptic SEgmentation, which unifies pre-trained text-image diffusion and discriminative models to perform open-vocabulary panoptic segmentation. Text-to-image diffusion models have shown the remarkable capability of generating high-quality images with diverse open-vocabulary language descriptions. This demonstrates that their internal representation space is highly correlated with open concepts in the real world. Text-image discriminative models like CLIP, on the other hand, are good at classifying images into open-vocabulary labels. We propose to leverage the frozen representation of both these models to perform panoptic segmentation of any category in the wild. Our approach outperforms the previous state of the art by significant margins on both open-vocabulary panoptic and semantic segmentation tasks. In particular, with COCO training only, our method achieves 23.4 PQ and 30.0 mIoU on the ADE20K dataset, with 8.3 PQ and 7.9 mIoU absolute improvement over the previous state-of-the-art. Project page is available at \url{https://jerryxu.net/ODISE}.
comment: CVPR 2022. Project page: https://jerryxu.net/ODISE
☆ Multilevel Diffusion: Infinite Dimensional Score-Based Diffusion Models for Image Generation
Score-based diffusion models (SBDM) have recently emerged as state-of-the-art approaches for image generation. Existing SBDMs are typically formulated in a finite-dimensional setting, where images are considered as tensors of a finite size. This papers develops SBDMs in the infinite-dimensional setting, that is, we model the training data as functions supported on a rectangular domain. Besides the quest for generating images at ever higher resolution our primary motivation is to create a well-posed infinite-dimensional learning problem so that we can discretize it consistently on multiple resolution levels. We thereby hope to obtain diffusion models that generalize across different resolution levels and improve the efficiency of the training process. We demonstrate how to overcome two shortcomings of current SBDM approaches in the infinite-dimensional setting. First, we modify the forward process to ensure that the latent distribution is well-defined in the infinite-dimensional setting using the notion of trace class operators. Second, we illustrate that approximating the score function with an operator network, in our case Fourier neural operators (FNOs), is beneficial for multilevel training. After deriving the forward and reverse process in the infinite-dimensional setting, we show their well-posedness, derive adequate discretizations, and investigate the role of the latent distributions. We provide first promising numerical results on two datasets, MNIST and material structures. In particular, we show that multilevel training is feasible within this framework.
☆ FastFill: Efficient Compatible Model Update
In many retrieval systems the original high dimensional data (e.g., images) is mapped to a lower dimensional feature through a learned embedding model. The task of retrieving the most similar data from a gallery set to a given query data is performed through a similarity comparison on features. When the embedding model is updated, it might produce features that are not comparable/compatible with features already in the gallery computed with the old model. Subsequently, all features in the gallery need to be re-computed using the new embedding model -- a computationally expensive process called backfilling. Recently, compatible representation learning methods have been proposed to avoid backfilling. Despite their relative success, there is an inherent trade-off between the new model performance and its compatibility with the old model. In this work, we introduce FastFill: a compatible model update process using feature alignment and policy based partial backfilling to promptly elevate retrieval performance. We show that previous backfilling strategies suffer from decreased performance and demonstrate the importance of both the training objective and the ordering in online partial backfilling. We propose a new training method for feature alignment between old and new embedding models using uncertainty estimation. Compared to previous works, we obtain significantly improved backfilling results on a variety of datasets: mAP on ImageNet (+4.4\%), Places-365 (+2.7\%), and VGG-Face2 (+1.3\%). Further, we demonstrate that when updating a biased model with FastFill, the minority subgroup accuracy gap promptly vanishes with a small fraction of partial backfilling.
comment: To appear in The Eleventh International Conference on Learning Representations
☆ Video-P2P: Video Editing with Cross-attention Control
This paper presents Video-P2P, a novel framework for real-world video editing with cross-attention control. While attention control has proven effective for image editing with pre-trained image generation models, there are currently no large-scale video generation models publicly available. Video-P2P addresses this limitation by adapting an image generation diffusion model to complete various video editing tasks. Specifically, we propose to first tune a Text-to-Set (T2S) model to complete an approximate inversion and then optimize a shared unconditional embedding to achieve accurate video inversion with a small memory cost. For attention control, we introduce a novel decoupled-guidance strategy, which uses different guidance strategies for the source and target prompts. The optimized unconditional embedding for the source prompt improves reconstruction ability, while an initialized unconditional embedding for the target prompt enhances editability. Incorporating the attention maps of these two branches enables detailed editing. These technical designs enable various text-driven editing applications, including word swap, prompt refinement, and attention re-weighting. Video-P2P works well on real-world videos for generating new characters while optimally preserving their original poses and scenes. It significantly outperforms previous approaches.
comment: 10 pages, 9 figures. Project page: https://video-p2p.github.io/
☆ Multimodal Parameter-Efficient Few-Shot Class Incremental Learning
Few-Shot Class Incremental Learning (FSCIL) is a challenging continual learning task, where limited training examples are available during several learning sessions. To succeed in this task, it is necessary to avoid over-fitting new classes caused by biased distributions in the few-shot training sets. The general approach to address this issue involves enhancing the representational capability of a pre-defined backbone architecture by adding special modules for backward compatibility with older classes. However, this approach has not yet solved the dilemma of ensuring high classification accuracy over time while reducing the gap between the performance obtained on larger training sets and the smaller ones. In this work, we propose an alternative approach called Continual Parameter-Efficient CLIP (CPE-CLIP) to reduce the loss of information between different learning sessions. Instead of adapting additional modules to address information loss, we leverage the vast knowledge acquired by CLIP in large-scale pre-training and its effectiveness in generalizing to new concepts. Our approach is multimodal and parameter-efficient, relying on learnable prompts for both the language and vision encoders to enable transfer learning across sessions. We also introduce prompt regularization to improve performance and prevent forgetting. Our experimental results demonstrate that CPE-CLIP significantly improves FSCIL performance compared to state-of-the-art proposals while also drastically reducing the number of learnable parameters and training costs.
☆ CLIP-FO3D: Learning Free Open-world 3D Scene Representations from 2D Dense CLIP
Training a 3D scene understanding model requires complicated human annotations, which are laborious to collect and result in a model only encoding close-set object semantics. In contrast, vision-language pre-training models (e.g., CLIP) have shown remarkable open-world reasoning properties. To this end, we propose directly transferring CLIP's feature space to 3D scene understanding model without any form of supervision. We first modify CLIP's input and forwarding process so that it can be adapted to extract dense pixel features for 3D scene contents. We then project multi-view image features to the point cloud and train a 3D scene understanding model with feature distillation. Without any annotations or additional training, our model achieves promising annotation-free semantic segmentation results on open-vocabulary semantics and long-tailed concepts. Besides, serving as a cross-modal pre-training framework, our method can be used to improve data efficiency during fine-tuning. Our model outperforms previous SOTA methods in various zero-shot and data-efficient learning benchmarks. Most importantly, our model successfully inherits CLIP's rich-structured knowledge, allowing 3D scene understanding models to recognize not only object concepts but also open-world semantics.
☆ Advancing Direct Convolution using Convolution Slicing Optimization and ISA Extensions
Convolution is one of the most computationally intensive operations that must be performed for machine-learning model inference. A traditional approach to compute convolutions is known as the Im2Col + BLAS method. This paper proposes SConv: a direct-convolution algorithm based on a MLIR/LLVM code-generation toolchain that can be integrated into machine-learning compilers . This algorithm introduces: (a) Convolution Slicing Analysis (CSA) - a convolution-specific 3D cache-blocking analysis pass that focuses on tile reuse over the cache hierarchy; (b) Convolution Slicing Optimization (CSO) - a code-generation pass that uses CSA to generate a tiled direct-convolution macro-kernel; and (c) Vector-Based Packing (VBP) - an architecture-specific optimized input-tensor packing solution based on vector-register shift instructions for convolutions with unitary stride. Experiments conducted on 393 convolutions from full ONNX-MLIR machine-learning models indicate that the elimination of the Im2Col transformation and the use of fast packing routines result in a total packing time reduction, on full model inference, of 2.0x - 3.9x on Intel x86 and 3.6x - 7.2x on IBM POWER10. The speed-up over an Im2Col + BLAS method based on current BLAS implementations for end-to-end machine-learning model inference is in the range of 9% - 25% for Intel x86 and 10% - 42% for IBM POWER10 architectures. The total convolution speedup for model inference is 12% - 27% on Intel x86 and 26% - 46% on IBM POWER10. SConv also outperforms BLAS GEMM, when computing pointwise convolutions, in more than 83% of the 219 tested instances.
comment: 15 pages, 11 figures
☆ SoftMatch Distance: A Novel Distance for Weakly-Supervised Trend Change Detection in Bi-Temporal Images
General change detection (GCD) and semantic change detection (SCD) are common methods for identifying changes and distinguishing object categories involved in those changes, respectively. However, the binary changes provided by GCD is often not practical enough, while annotating semantic labels for training SCD models is very expensive. Therefore, there is a novel solution that intuitively dividing changes into three trends (``appear'', ``disappear'' and ``transform'') instead of semantic categories, named it trend change detection (TCD) in this paper. It offers more detailed change information than GCD, while requiring less manual annotation cost than SCD. However, there are limited public data sets with specific trend labels to support TCD application. To address this issue, we propose a softmatch distance which is used to construct a weakly-supervised TCD branch in a simple GCD model, using GCD labels instead of TCD label for training. Furthermore, a strategic approach is presented to successfully explore and extract background information, which is crucial for the weakly-supervised TCD task. The experiment results on four public data sets are highly encouraging, which demonstrates the effectiveness of our proposed model.
☆ Towards Trust of Explainable AI in Thyroid Nodule Diagnosis AAAI 2023
The ability to explain the prediction of deep learning models to end-users is an important feature to leverage the power of artificial intelligence (AI) for the medical decision-making process, which is usually considered non-transparent and challenging to comprehend. In this paper, we apply state-of-the-art eXplainable artificial intelligence (XAI) methods to explain the prediction of the black-box AI models in the thyroid nodule diagnosis application. We propose new statistic-based XAI methods, namely Kernel Density Estimation and Density map, to explain the case of no nodule detected. XAI methods' performances are considered under a qualitative and quantitative comparison as feedback to improve the data quality and the model performance. Finally, we survey to assess doctors' and patients' trust in XAI explanations of the model's decisions on thyroid nodule images.
comment: Accepted by AAAI 2023 The 7th International Workshop on Health Intelligence (W3PHIAI-23)
☆ Medical Waste Sorting: a computer vision approach for assisted primary sorting
Medical waste, i.e. waste produced during medical activities in hospitals, clinics and laboratories, represents hazardous waste whose management involves special care and high costs. However, this kind of waste contains a significant fraction of highly valued materials that can enter a circular economy process. To this end, in this paper, we propose a computer vision approach for assisting in the primary sorting of medical waste. The feasibility of our approach is demonstrated on a representative dataset we collected and made available to the community, with which we have trained a model that achieves 100\% accuracy, and a new dataset on which the trained model exhibits good generalization.
comment: Submitted to IWCIM 2023
☆ DiM: Distilling Dataset into Generative Model
Dataset distillation reduces the network training cost by synthesizing small and informative datasets from large-scale ones. Despite the success of the recent dataset distillation algorithms, three drawbacks still limit their wider application: i). the synthetic images perform poorly on large architectures; ii). they need to be re-optimized when the distillation ratio changes; iii). the limited diversity restricts the performance when the distillation ratio is large. In this paper, we propose a novel distillation scheme to \textbf{D}istill information of large train sets \textbf{i}nto generative \textbf{M}odels, named DiM. Specifically, DiM learns to use a generative model to store the information of the target dataset. During the distillation phase, we minimize the differences in logits predicted by a models pool between real and generated images. At the deployment stage, the generative model synthesizes various training samples from random noises on the fly. Due to the simple yet effective designs, the trained DiM can be directly applied to different distillation ratios and large architectures without extra cost. We validate the proposed DiM across 4 datasets and achieve state-of-the-art results on all of them. To the best of our knowledge, we are the first to achieve higher accuracy on complex architectures than simple ones, such as 75.1\% with ResNet-18 and 72.6\% with ConvNet-3 on ten images per class of CIFAR-10. Besides, DiM outperforms previous methods with 10\% $\sim$ 22\% when images per class are 1 and 10 on the SVHN dataset.
comment: Distilling datasets into generative models
☆ VOLTA: an Environment-Aware Contrastive Cell Representation Learning for Histopathology
In clinical practice, many diagnosis tasks rely on the identification of cells in histopathology images. While supervised machine learning techniques require labels, providing manual cell annotations is time-consuming due to the large number of cells. In this paper, we propose a self-supervised framework (VOLTA) for cell representation learning in histopathology images using a novel technique that accounts for the cell's mutual relationship with its environment for improved cell representations. We subjected our model to extensive experiments on the data collected from multiple institutions around the world comprising of over 700,000 cells, four cancer types, and cell types ranging from three to six categories for each dataset. The results show that our model outperforms the state-of-the-art models in cell representation learning. To showcase the potential power of our proposed framework, we applied VOLTA to ovarian and endometrial cancers with very small sample sizes (10-20 samples) and demonstrated that our cell representations can be utilized to identify the known histotypes of ovarian cancer and provide novel insights that link histopathology and molecular subtypes of endometrial cancer. Unlike supervised deep learning models that require large sample sizes for training, we provide a framework that can empower new discoveries without any annotation data in situations where sample sizes are limited.
☆ Enhancing Low-resolution Face Recognition with Feature Similarity Knowledge Distillation
In this study, we introduce a feature knowledge distillation framework to improve low-resolution (LR) face recognition performance using knowledge obtained from high-resolution (HR) images. The proposed framework transfers informative features from an HR-trained network to an LR-trained network by reducing the distance between them. A cosine similarity measure was employed as a distance metric to effectively align the HR and LR features. This approach differs from conventional knowledge distillation frameworks, which use the L_p distance metrics and offer the advantage of converging well when reducing the distance between features of different resolutions. Our framework achieved a 3% improvement over the previous state-of-the-art method on the AgeDB-30 benchmark without bells and whistles, while maintaining a strong performance on HR images. The effectiveness of cosine similarity as a distance metric was validated through statistical analysis, making our approach a promising solution for real-world applications in which LR images are frequently encountered. The code and pretrained models will be publicly available on GitHub.
☆ Visual ChatGPT: Talking, Drawing and Editing with Visual Foundation Models
ChatGPT is attracting a cross-field interest as it provides a language interface with remarkable conversational competency and reasoning capabilities across many domains. However, since ChatGPT is trained with languages, it is currently not capable of processing or generating images from the visual world. At the same time, Visual Foundation Models, such as Visual Transformers or Stable Diffusion, although showing great visual understanding and generation capabilities, they are only experts on specific tasks with one-round fixed inputs and outputs. To this end, We build a system called \textbf{Visual ChatGPT}, incorporating different Visual Foundation Models, to enable the user to interact with ChatGPT by 1) sending and receiving not only languages but also images 2) providing complex visual questions or visual editing instructions that require the collaboration of multiple AI models with multi-steps. 3) providing feedback and asking for corrected results. We design a series of prompts to inject the visual model information into ChatGPT, considering models of multiple inputs/outputs and models that require visual feedback. Experiments show that Visual ChatGPT opens the door to investigating the visual roles of ChatGPT with the help of Visual Foundation Models. Our system is publicly available at \url{https://github.com/microsoft/visual-chatgpt}.
☆ EvConv: Fast CNN Inference on Event Camera Inputs For High-Speed Robot Perception
Event cameras capture visual information with a high temporal resolution and a wide dynamic range. This enables capturing visual information at fine time granularities (e.g., microseconds) in rapidly changing environments. This makes event cameras highly useful for high-speed robotics tasks involving rapid motion, such as high-speed perception, object tracking, and control. However, convolutional neural network inference on event camera streams cannot currently perform real-time inference at the high speeds at which event cameras operate - current CNN inference times are typically closer in order of magnitude to the frame rates of regular frame-based cameras. Real-time inference at event camera rates is necessary to fully leverage the high frequency and high temporal resolution that event cameras offer. This paper presents EvConv, a new approach to enable fast inference on CNNs for inputs from event cameras. We observe that consecutive inputs to the CNN from an event camera have only small differences between them. Thus, we propose to perform inference on the difference between consecutive input tensors, or the increment. This enables a significant reduction in the number of floating-point operations required (and thus the inference latency) because increments are very sparse. We design EvConv to leverage the irregular sparsity in increments from event cameras and to retain the sparsity of these increments across all layers of the network. We demonstrate a reduction in the number of floating operations required in the forward pass by up to 98%. We also demonstrate a speedup of up to 1.6X for inference using CNNs for tasks such as depth estimation, object recognition, and optical flow estimation, with almost no loss in accuracy.
☆ STPDnet: Spatial-temporal convolutional primal dual network for dynamic PET image reconstruction
Dynamic positron emission tomography (dPET) image reconstruction is extremely challenging due to the limited counts received in individual frame. In this paper, we propose a spatial-temporal convolutional primal dual network (STPDnet) for dynamic PET image reconstruction. Both spatial and temporal correlations are encoded by 3D convolution operators. The physical projection of PET is embedded in the iterative learning process of the network, which provides the physical constraints and enhances interpretability. The experiments of real rat scan data have shown that the proposed method can achieve substantial noise reduction in both temporal and spatial domains and outperform the maximum likelihood expectation maximization (MLEM), spatial-temporal kernel method (KEM-ST), DeepPET and Learned Primal Dual (LPD).
comment: ISBI2023 accepted
☆ Centroid-centered Modeling for Efficient Vision Transformer Pre-training
Masked Image Modeling (MIM) is a new self-supervised vision pre-training paradigm using Vision Transformer (ViT). Previous works can be pixel-based or token-based, using original pixels or discrete visual tokens from parametric tokenizer models, respectively. Our proposed approach, \textbf{CCViT}, leverages k-means clustering to obtain centroids for image modeling without supervised training of tokenizer model. The centroids represent patch pixels and index tokens and have the property of local invariance. Non-parametric centroid tokenizer only takes seconds to create and is faster for token inference. Specifically, we adopt patch masking and centroid replacement strategies to construct corrupted inputs, and two stacked encoder blocks to predict corrupted patch tokens and reconstruct original patch pixels. Experiments show that the ViT-B model with only 300 epochs achieves 84.3\% top-1 accuracy on ImageNet-1K classification and 51.6\% on ADE20K semantic segmentation. Our approach achieves competitive results with BEiTv2 without distillation training from other models and outperforms other methods such as MAE.
☆ DULDA: Dual-domain Unsupervised Learned Descent Algorithm for PET image reconstruction
Deep learning based PET image reconstruction methods have achieved promising results recently. However, most of these methods follow a supervised learning paradigm, which rely heavily on the availability of high-quality training labels. In particular, the long scanning time required and high radiation exposure associated with PET scans make obtaining this labels impractical. In this paper, we propose a dual-domain unsupervised PET image reconstruction method based on learned decent algorithm, which reconstructs high-quality PET images from sinograms without the need for image labels. Specifically, we unroll the proximal gradient method with a learnable l2,1 norm for PET image reconstruction problem. The training is unsupervised, using measurement domain loss based on deep image prior as well as image domain loss based on rotation equivariance property. The experimental results domonstrate the superior performance of proposed method compared with maximum likelihood expectation maximazation (MLEM), total-variation regularized EM (EM-TV) and deep image prior based method (DIP).
☆ Aberration-Aware Depth-from-Focus
Computer vision methods for depth estimation usually use simple camera models with idealized optics. For modern machine learning approaches, this creates an issue when attempting to train deep networks with simulated data, especially for focus-sensitive tasks like Depth-from-Focus. In this work, we investigate the domain gap caused by off-axis aberrations that will affect the decision of the best-focused frame in a focal stack. We then explore bridging this domain gap through aberration-aware training (AAT). Our approach involves a lightweight network that models lens aberrations at different positions and focus distances, which is then integrated into the conventional network training pipeline. We evaluate the generality of pretrained models on both synthetic and real-world data. Our experimental results demonstrate that the proposed AAT scheme can improve depth estimation accuracy without fine-tuning the model or modifying the network architecture.
comment: Considering optical aberrations during network training can improve the generalizability
Transformer-based Image Generation from Scene Graphs
Graph-structured scene descriptions can be efficiently used in generative models to control the composition of the generated image. Previous approaches are based on the combination of graph convolutional networks and adversarial methods for layout prediction and image generation, respectively. In this work, we show how employing multi-head attention to encode the graph information, as well as using a transformer-based model in the latent space for image generation can improve the quality of the sampled data, without the need to employ adversarial models with the subsequent advantage in terms of training stability. The proposed approach, specifically, is entirely based on transformer architectures both for encoding scene graphs into intermediate object layouts and for decoding these layouts into images, passing through a lower dimensional space learned by a vector-quantized variational autoencoder. Our approach shows an improved image quality with respect to state-of-the-art methods as well as a higher degree of diversity among multiple generations from the same scene graph. We evaluate our approach on three public datasets: Visual Genome, COCO, and CLEVR. We achieve an Inception Score of 13.7 and 12.8, and an FID of 52.3 and 60.3, on COCO and Visual Genome, respectively. We perform ablation studies on our contributions to assess the impact of each component. Code is available at https://github.com/perceivelab/trf-sg2im
☆ Simple and Efficient Confidence Score for Grading Whole Slide Images
Grading precancerous lesions on whole slide images is a challenging task: the continuous space of morphological phenotypes makes clear-cut decisions between different grades often difficult, leading to low inter- and intra-rater agreements. More and more Artificial Intelligence (AI) algorithms are developed to help pathologists perform and standardize their diagnosis. However, those models can render their prediction without consideration of the ambiguity of the classes and can fail without notice which prevent their wider acceptance in a clinical context. In this paper, we propose a new score to measure the confidence of AI models in grading tasks. Our confidence score is specifically adapted to ordinal output variables, is versatile and does not require extra training or additional inferences nor particular architecture changes. Comparison to other popular techniques such as Monte Carlo Dropout and deep ensembles shows that our method provides state-of-the art results, while being simpler, more versatile and less computationally intensive. The score is also easily interpretable and consistent with real life hesitations of pathologists. We show that the score is capable of accurately identifying mispredicted slides and that accuracy for high confidence decisions is significantly higher than for low-confidence decisions (gap in AUC of 17.1% on the test set). We believe that the proposed confidence score could be leveraged by pathologists directly in their workflow and assist them on difficult tasks such as grading precancerous lesions.
☆ Learning Enhancement From Degradation: A Diffusion Model For Fundus Image Enhancement
The quality of a fundus image can be compromised by numerous factors, many of which are challenging to be appropriately and mathematically modeled. In this paper, we introduce a novel diffusion model based framework, named Learning Enhancement from Degradation (LED), for enhancing fundus images. Specifically, we first adopt a data-driven degradation framework to learn degradation mappings from unpaired high-quality to low-quality images. We then apply a conditional diffusion model to learn the inverse enhancement process in a paired manner. The proposed LED is able to output enhancement results that maintain clinically important features with better clarity. Moreover, in the inference phase, LED can be easily and effectively integrated with any existing fundus image enhancement framework. We evaluate the proposed LED on several downstream tasks with respect to various clinically-relevant metrics, successfully demonstrating its superiority over existing state-of-the-art methods both quantitatively and qualitatively. The source code is available at https://github.com/QtacierP/LED.
☆ Point Cloud Classification Using Content-based Transformer via Clustering in Feature Space
Recently, there have been some attempts of Transformer in 3D point cloud classification. In order to reduce computations, most existing methods focus on local spatial attention, but ignore their content and fail to establish relationships between distant but relevant points. To overcome the limitation of local spatial attention, we propose a point content-based Transformer architecture, called PointConT for short. It exploits the locality of points in the feature space (content-based), which clusters the sampled points with similar features into the same class and computes the self-attention within each class, thus enabling an effective trade-off between capturing long-range dependencies and computational complexity. We further introduce an Inception feature aggregator for point cloud classification, which uses parallel structures to aggregate high-frequency and low-frequency information in each branch separately. Extensive experiments show that our PointConT model achieves a remarkable performance on point cloud shape classification. Especially, our method exhibits 90.3% Top-1 accuracy on the hardest setting of ScanObjectNN. Source code of this paper is available at https://github.com/yahuiliu99/PointConT.
comment: This paper is accepted to IEEE/CAA Journal of Automatica Sinica (JAS)
☆ Structure-aware registration network for liver DCE-CT images
Image registration of liver dynamic contrast-enhanced computed tomography (DCE-CT) is crucial for diagnosis and image-guided surgical planning of liver cancer. However, intensity variations due to the flow of contrast agents combined with complex spatial motion induced by respiration brings great challenge to existing intensity-based registration methods. To address these problems, we propose a novel structure-aware registration method by incorporating structural information of related organs with segmentation-guided deep registration network. Existing segmentation-guided registration methods only focus on volumetric registration inside the paired organ segmentations, ignoring the inherent attributes of their anatomical structures. In addition, such paired organ segmentations are not always available in DCE-CT images due to the flow of contrast agents. Different from existing segmentation-guided registration methods, our proposed method extracts structural information in hierarchical geometric perspectives of line and surface. Then, according to the extracted structural information, structure-aware constraints are constructed and imposed on the forward and backward deformation field simultaneously. In this way, all available organ segmentations, including unpaired ones, can be fully utilized to avoid the side effect of contrast agent and preserve the topology of organs during registration. Extensive experiments on an in-house liver DCE-CT dataset and a public LiTS dataset show that our proposed method can achieve higher registration accuracy and preserve anatomical structure more effectively than state-of-the-art methods.
☆ FCN+: Global Receptive Convolution Makes FCN Great Again
Fully convolutional network (FCN) is a seminal work for semantic segmentation. However, due to its limited receptive field, FCN cannot effectively capture global context information which is vital for semantic segmentation. As a result, it is beaten by state-of-the-art methods which leverage different filter sizes for larger receptive fields. However, such a strategy usually introduces more parameters and increases the computational cost. In this paper, we propose a novel global receptive convolution (GRC) to effectively increase the receptive field of FCN for context information extraction, which results in an improved FCN termed FCN+. The GRC provides global receptive field for convolution without introducing any extra learnable parameters. The motivation of GRC is that different channels of a convolutional filter can have different grid sampling locations across the whole input feature map. Specifically, the GRC first divides the channels of the filter into two groups. The grid sampling locations of the first group are shifted to different spatial coordinates across the whole feature map, according to their channel indexes. This can help the convolutional filter capture the global context information. The grid sampling location of the second group remains unchanged to keep the original location information. Convolving using these two groups, the GRC can integrate the global context into the original location information of each pixel for better dense prediction results. With the GRC built in, FCN+ can achieve comparable performance to state-of-the-art methods for semantic segmentation tasks, as verified on PASCAL VOC 2012, Cityscapes, and ADE20K.
☆ A Prompt Log Analysis of Text-to-Image Generation Systems
Recent developments in diffusion models have unleashed the astonishing capabilities of text-to-image generation systems to synthesize high-quality images that are faithful to a given reference text, known as a "prompt." These systems, once released to the public, have immediately received tons of attention from researchers, creators, and common users. Despite the plenty of efforts to improve the underneath generative models, there is limited work on understanding the information needs of the real users of these systems, e.g., by investigating the prompts the users input at scale. In this paper, we take the initiative to conduct a comprehensive analysis of large-scale prompt logs collected from multiple text-to-image generation systems. Our work is analogous to analyzing the query log of Web search engines, a line of work that has made critical contributions to the glory of the Web search industry and research. We analyze over two million user-input prompts submitted to three popular text-to-image systems at scale. Compared to Web search queries, text-to-image prompts are significantly longer, often organized into unique structures, and present different categories of information needs. Users tend to make more edits within creation sessions, showing remarkable exploratory patterns. Our findings provide concrete implications on how to improve text-to-image generation systems for creation purposes.
☆ Robustness Evaluation in Hand Pose Estimation Models using Metamorphic Testing
Hand pose estimation (HPE) is a task that predicts and describes the hand poses from images or video frames. When HPE models estimate hand poses captured in a laboratory or under controlled environments, they normally deliver good performance. However, the real-world environment is complex, and various uncertainties may happen, which could degrade the performance of HPE models. For example, the hands could be occluded, the visibility of hands could be reduced by imperfect exposure rate, and the contour of hands prone to be blurred during fast hand movements. In this work, we adopt metamorphic testing to evaluate the robustness of HPE models and provide suggestions on the choice of HPE models for different applications. The robustness evaluation was conducted on four state-of-the-art models, namely MediaPipe hands, OpenPose, BodyHands, and NSRM hand. We found that on average more than 80\% of the hands could not be identified by BodyHands, and at least 50\% of hands could not be identified by MediaPipe hands when diagonal motion blur is introduced, while an average of more than 50\% of strongly underexposed hands could not be correctly estimated by NSRM hand. Similarly, applying occlusions on only four hand joints will also largely degrade the performance of these models. The experimental results show that occlusions, illumination variations, and motion blur are the main obstacles to the performance of existing HPE models. These findings may pave the way for researchers to improve the performance and robustness of hand pose estimation models and their applications.
comment: Accepted at 2023 8th International Workshop on Metamorphic Testing, 8 pages
☆ Scene Matters: Model-based Deep Video Compression
Video compression has always been a popular research area, where many traditional and deep video compression methods have been proposed. These methods typically rely on signal prediction theory to enhance compression performance by designing high efficient intra and inter prediction strategies and compressing video frames one by one. In this paper, we propose a novel model-based video compression (MVC) framework that regards scenes as the fundamental units for video sequences. Our proposed MVC directly models the intensity variation of the entire video sequence in one scene, seeking non-redundant representations instead of reducing redundancy through spatio-temporal predictions. To achieve this, we employ implicit neural representation (INR) as our basic modeling architecture. To improve the efficiency of video modeling, we first propose context-related spatial positional embedding (CRSPE) and frequency domain supervision (FDS) in spatial context enhancement. For temporal correlation capturing, we design the scene flow constrain mechanism (SFCM) and temporal contrastive loss (TCL). Extensive experimental results demonstrate that our method achieves up to a 20\% bitrate reduction compared to the latest video coding standard H.266 and is more efficient in decoding than existing video coding strategies.
☆ RADAM: Texture Recognition through Randomized Aggregated Encoding of Deep Activation Maps
Texture analysis is a classical yet challenging task in computer vision for which deep neural networks are actively being applied. Most approaches are based on building feature aggregation modules around a pre-trained backbone and then fine-tuning the new architecture on specific texture recognition tasks. Here we propose a new method named \textbf{R}andom encoding of \textbf{A}ggregated \textbf{D}eep \textbf{A}ctivation \textbf{M}aps (RADAM) which extracts rich texture representations without ever changing the backbone. The technique consists of encoding the output at different depths of a pre-trained deep convolutional network using a Randomized Autoencoder (RAE). The RAE is trained locally to each image using a closed-form solution, and its decoder weights are used to compose a 1-dimensional texture representation that is fed into a linear SVM. This means that no fine-tuning or backpropagation is needed. We explore RADAM on several texture benchmarks and achieve state-of-the-art results with different computational budgets. Our results suggest that pre-trained backbones may not require additional fine-tuning for texture recognition if their learned representations are better encoded.
comment: 17 pages, 3 figures, submitted to peer-review journal
☆ Continuity-Aware Latent Interframe Information Mining for Reliable UAV Tracking ICRA
Unmanned aerial vehicle (UAV) tracking is crucial for autonomous navigation and has broad applications in robotic automation fields. However, reliable UAV tracking remains a challenging task due to various difficulties like frequent occlusion and aspect ratio change. Additionally, most of the existing work mainly focuses on explicit information to improve tracking performance, ignoring potential interframe connections. To address the above issues, this work proposes a novel framework with continuity-aware latent interframe information mining for reliable UAV tracking, i.e., ClimRT. Specifically, a new efficient continuity-aware latent interframe information mining network (ClimNet) is proposed for UAV tracking, which can generate highly-effective latent frame between two adjacent frames. Besides, a novel location-continuity Transformer (LCT) is designed to fully explore continuity-aware spatial-temporal information, thereby markedly enhancing UAV tracking. Extensive qualitative and quantitative experiments on three authoritative aerial benchmarks strongly validate the robustness and reliability of ClimRT in UAV tracking performance. Furthermore, real-world tests on the aerial platform validate its practicability and effectiveness. The code and demo materials are released at https://github.com/vision4robotics/ClimRT.
comment: 2023 IEEE International Conference on Robotics and Automation (ICRA)
☆ FastSurf: Fast Neural RGB-D Surface Reconstruction using Per-Frame Intrinsic Refinement and TSDF Fusion Prior Learning
We introduce FastSurf, an accelerated neural radiance field (NeRF) framework that incorporates depth information for 3D reconstruction. A dense feature grid and shallow multi-layer perceptron are used for fast and accurate surface optimization of the entire scene. Our per-frame intrinsic refinement scheme corrects the frame-specific errors that cannot be handled by global optimization. Furthermore, FastSurf utilizes a classical real-time 3D surface reconstruction method, the truncated signed distance field (TSDF) Fusion, as prior knowledge to pretrain the feature grid to accelerate the training. The quantitative and qualitative experiments comparing the performances of FastSurf against prior work indicate that our method is capable of quickly and accurately reconstructing a scene with high-frequency details. We also demonstrate the effectiveness of our per-frame intrinsic refinement and TSDF Fusion prior learning techniques via an ablation study.
☆ Radio astronomical images object detection and segmentation: A benchmark on deep learning methods
In recent years, deep learning has been successfully applied in various scientific domains. Following these promising results and performances, it has recently also started being evaluated in the domain of radio astronomy. In particular, since radio astronomy is entering the Big Data era, with the advent of the largest telescope in the world - the Square Kilometre Array (SKA), the task of automatic object detection and instance segmentation is crucial for source finding and analysis. In this work, we explore the performance of the most affirmed deep learning approaches, applied to astronomical images obtained by radio interferometric instrumentation, to solve the task of automatic source detection. This is carried out by applying models designed to accomplish two different kinds of tasks: object detection and semantic segmentation. The goal is to provide an overview of existing techniques, in terms of prediction performance and computational efficiency, to scientists in the astrophysics community who would like to employ machine learning in their research.
☆ Immune Defense: A Novel Adversarial Defense Mechanism for Preventing the Generation of Adversarial Examples
The vulnerability of Deep Neural Networks (DNNs) to adversarial examples has been confirmed. Existing adversarial defenses primarily aim at preventing adversarial examples from attacking DNNs successfully, rather than preventing their generation. If the generation of adversarial examples is unregulated, images within reach are no longer secure and pose a threat to non-robust DNNs. Although gradient obfuscation attempts to address this issue, it has been shown to be circumventable. Therefore, we propose a novel adversarial defense mechanism, which is referred to as immune defense and is the example-based pre-defense. This mechanism applies carefully designed quasi-imperceptible perturbations to the raw images to prevent the generation of adversarial examples for the raw images, and thereby protecting both images and DNNs. These perturbed images are referred to as Immune Examples (IEs). In the white-box immune defense, we provide a gradient-based and an optimization-based approach, respectively. Additionally, the more complex black-box immune defense is taken into consideration. We propose Masked Gradient Sign Descent (MGSD) to reduce approximation error and stabilize the update to improve the transferability of IEs and thereby ensure their effectiveness against black-box adversarial attacks. The experimental results demonstrate that the optimization-based approach has superior performance and better visual quality in white-box immune defense. In contrast, the gradient-based approach has stronger transferability and the proposed MGSD significantly improve the transferability of baselines.
☆ Exploiting the Textual Potential from Vision-Language Pre-training for Text-based Person Search
Text-based Person Search (TPS), is targeted on retrieving pedestrians to match text descriptions instead of query images. Recent Vision-Language Pre-training (VLP) models can bring transferable knowledge to downstream TPS tasks, resulting in more efficient performance gains. However, existing TPS methods improved by VLP only utilize pre-trained visual encoders, neglecting the corresponding textual representation and breaking the significant modality alignment learned from large-scale pre-training. In this paper, we explore the full utilization of textual potential from VLP in TPS tasks. We build on the proposed VLP-TPS baseline model, which is the first TPS model with both pre-trained modalities. We propose the Multi-Integrity Description Constraints (MIDC) to enhance the robustness of the textual modality by incorporating different components of fine-grained corpus during training. Inspired by the prompt approach for zero-shot classification with VLP models, we propose the Dynamic Attribute Prompt (DAP) to provide a unified corpus of fine-grained attributes as language hints for the image modality. Extensive experiments show that our proposed TPS framework achieves state-of-the-art performance, exceeding the previous best method by a margin.
comment: 10 pages, 6 figures
☆ DANet: Density Adaptive Convolutional Network with Interactive Attention for 3D Point Clouds
Local features and contextual dependencies are crucial for 3D point cloud analysis. Many works have been devoted to designing better local convolutional kernels that exploit the contextual dependencies. However, current point convolutions lack robustness to varying point cloud density. Moreover, contextual modeling is dominated by non-local or self-attention models which are computationally expensive. To solve these problems, we propose density adaptive convolution, coined DAConv. The key idea is to adaptively learn the convolutional weights from geometric connections obtained from the point density and position. To extract precise context dependencies with fewer computations, we propose an interactive attention module (IAM) that embeds spatial information into channel attention along different spatial directions. DAConv and IAM are integrated in a hierarchical network architecture to achieve local density and contextual direction-aware learning for point cloud analysis. Experiments show that DAConv is significantly more robust to point density compared to existing methods and extensive comparisons on challenging 3D point cloud datasets show that our network achieves state-of-the-art classification results of 93.6% on ModelNet40, competitive semantic segmentation results of 68.71% mIoU on S3DIS and part segmentation results of 86.7% mIoU on ShapeNet.
comment: 9
☆ Full Point Encoding for Local Feature Aggregation in 3D Point Clouds
Point cloud processing methods exploit local point features and global context through aggregation which does not explicity model the internal correlations between local and global features. To address this problem, we propose full point encoding which is applicable to convolution and transformer architectures. Specifically, we propose Full Point Convolution (FPConv) and Full Point Transformer (FPTransformer) architectures. The key idea is to adaptively learn the weights from local and global geometric connections, where the connections are established through local and global correlation functions respectively. FPConv and FPTransformer simultaneously model the local and global geometric relationships as well as their internal correlations, demonstrating strong generalization ability and high performance. FPConv is incorporated in classical hierarchical network architectures to achieve local and global shape-aware learning. In FPTransformer, we introduce full point position encoding in self-attention, that hierarchically encodes each point position in the global and local receptive field. We also propose a shape aware downsampling block which takes into account the local shape and the global context. Experimental comparison to existing methods on benchmark datasets show the efficacy of FPConv and FPTransformer for semantic segmentation, object detection, classification, and normal estimation tasks. In particular, we achieve state-of-the-art semantic segmentation results of 76% mIoU on S3DIS 6-fold and 72.2% on S3DIS Area5.
comment: 15
☆ RM-Depth: Unsupervised Learning of Recurrent Monocular Depth in Dynamic Scenes CVPR 2022
Unsupervised methods have showed promising results on monocular depth estimation. However, the training data must be captured in scenes without moving objects. To push the envelope of accuracy, recent methods tend to increase their model parameters. In this paper, an unsupervised learning framework is proposed to jointly predict monocular depth and complete 3D motion including the motions of moving objects and camera. (1) Recurrent modulation units are used to adaptively and iteratively fuse encoder and decoder features. This not only improves the single-image depth inference but also does not overspend model parameters. (2) Instead of using a single set of filters for upsampling, multiple sets of filters are devised for the residual upsampling. This facilitates the learning of edge-preserving filters and leads to the improved performance. (3) A warping-based network is used to estimate a motion field of moving objects without using semantic priors. This breaks down the requirement of scene rigidity and allows to use general videos for the unsupervised learning. The motion field is further regularized by an outlier-aware training loss. Despite the depth model just uses a single image in test time and 2.97M parameters, it achieves state-of-the-art results on the KITTI and Cityscapes benchmarks.
comment: Accepted to CVPR 2022 (paper is updated)
☆ Grasping Student: semi-supervised learning for robotic manipulation
Gathering real-world data from the robot quickly becomes a bottleneck when constructing a robot learning system for grasping. In this work, we design a semi-supervised grasping system that, on top of a small sample of robot experience, takes advantage of images of products to be picked, which are collected without any interactions with the robot. We validate our findings both in the simulation and in the real world. In the regime of a small number of robot training samples, taking advantage of the unlabeled data allows us to achieve performance at the level of 10-fold bigger dataset size used by the baseline. The code and datasets used in the paper will be released at https://github.com/nomagiclab/grasping-student.
☆ Loss-Curvature Matching for Dataset Selection and Condensation AISTATS
Training neural networks on a large dataset requires substantial computational costs. Dataset reduction selects or synthesizes data instances based on the large dataset, while minimizing the degradation in generalization performance from the full dataset. Existing methods utilize the neural network during the dataset reduction procedure, so the model parameter becomes important factor in preserving the performance after reduction. By depending upon the importance of parameters, this paper introduces a new reduction objective, coined LCMat, which Matches the Loss Curvatures of the original dataset and reduced dataset over the model parameter space, more than the parameter point. This new objective induces a better adaptation of the reduced dataset on the perturbed parameter region than the exact point matching. Particularly, we identify the worst case of the loss curvature gap from the local parameter region, and we derive the implementable upper bound of such worst-case with theoretical analyses. Our experiments on both coreset selection and condensation benchmarks illustrate that LCMat shows better generalization performances than existing baselines.
comment: 26th International Conference on Artificial Intelligence and Statistics (AISTATS)
☆ HyT-NAS: Hybrid Transformers Neural Architecture Search for Edge Devices
Vision Transformers have enabled recent attention-based Deep Learning (DL) architectures to achieve remarkable results in Computer Vision (CV) tasks. However, due to the extensive computational resources required, these architectures are rarely implemented on resource-constrained platforms. Current research investigates hybrid handcrafted convolution-based and attention-based models for CV tasks such as image classification and object detection. In this paper, we propose HyT-NAS, an efficient Hardware-aware Neural Architecture Search (HW-NAS) including hybrid architectures targeting vision tasks on tiny devices. HyT-NAS improves state-of-the-art HW-NAS by enriching the search space and enhancing the search strategy as well as the performance predictors. Our experiments show that HyT-NAS achieves a similar hypervolume with less than ~5x training evaluations. Our resulting architecture outperforms MLPerf MobileNetV1 by 6.3% accuracy improvement with 3.5x less number of parameters on Visual Wake Words.
comment: CODAI 2022 Workshop - Embedded System Week (ESWeek)
☆ A Light Weight Model for Active Speaker Detection CVPR 2023
Active speaker detection is a challenging task in audio-visual scenario understanding, which aims to detect who is speaking in one or more speakers scenarios. This task has received extensive attention as it is crucial in applications such as speaker diarization, speaker tracking, and automatic video editing. The existing studies try to improve performance by inputting multiple candidate information and designing complex models. Although these methods achieved outstanding performance, their high consumption of memory and computational power make them difficult to be applied in resource-limited scenarios. Therefore, we construct a lightweight active speaker detection architecture by reducing input candidates, splitting 2D and 3D convolutions for audio-visual feature extraction, and applying gated recurrent unit (GRU) with low computational complexity for cross-modal modeling. Experimental results on the AVA-ActiveSpeaker dataset show that our framework achieves competitive mAP performance (94.1% vs. 94.2%), while the resource costs are significantly lower than the state-of-the-art method, especially in model parameters (1.0M vs. 22.5M, about 23x) and FLOPs (0.6G vs. 2.6G, about 4x). In addition, our framework also performs well on the Columbia dataset showing good robustness. The code and model weights are available at https://github.com/Junhua-Liao/Light-ASD.
comment: Accepted by CVPR 2023
☆ A Message Passing Perspective on Learning Dynamics of Contrastive Learning ICLR 2023
In recent years, contrastive learning achieves impressive results on self-supervised visual representation learning, but there still lacks a rigorous understanding of its learning dynamics. In this paper, we show that if we cast a contrastive objective equivalently into the feature space, then its learning dynamics admits an interpretable form. Specifically, we show that its gradient descent corresponds to a specific message passing scheme on the corresponding augmentation graph. Based on this perspective, we theoretically characterize how contrastive learning gradually learns discriminative features with the alignment update and the uniformity update. Meanwhile, this perspective also establishes an intriguing connection between contrastive learning and Message Passing Graph Neural Networks (MP-GNNs). This connection not only provides a unified understanding of many techniques independently developed in each community, but also enables us to borrow techniques from MP-GNNs to design new contrastive learning variants, such as graph attention, graph rewiring, jumpy knowledge techniques, etc. We believe that our message passing perspective not only provides a new theoretical understanding of contrastive learning dynamics, but also bridges the two seemingly independent areas together, which could inspire more interleaving studies to benefit from each other. The code is available at https://github.com/PKU-ML/Message-Passing-Contrastive-Learning.
comment: ICLR 2023
Self-Supervised Learning for Group Equivariant Neural Networks
This paper proposes a method to construct pretext tasks for self-supervised learning on group equivariant neural networks. Group equivariant neural networks are the models whose structure is restricted to commute with the transformations on the input. Therefore, it is important to construct pretext tasks for self-supervised learning that do not contradict this equivariance. To ensure that training is consistent with the equivariance, we propose two concepts for self-supervised tasks: equivariant pretext labels and invariant contrastive loss. Equivariant pretext labels use a set of labels on which we can define the transformations that correspond to the input change. Invariant contrastive loss uses a modified contrastive loss that absorbs the effect of transformations on each input. Experiments on standard image recognition benchmarks demonstrate that the equivariant neural networks exploit the proposed equivariant self-supervised tasks.
comment: 12 pages, 4 figures
☆ FUSQA: Fetal Ultrasound Segmentation Quality Assessment
Deep learning models have been effective for various fetal ultrasound segmentation tasks. However, generalization to new unseen data has raised questions about their effectiveness for clinical adoption. Normally, a transition to new unseen data requires time-consuming and costly quality assurance processes to validate the segmentation performance post-transition. Segmentation quality assessment efforts have focused on natural images, where the problem has been typically formulated as a dice score regression task. In this paper, we propose a simplified Fetal Ultrasound Segmentation Quality Assessment (FUSQA) model to tackle the segmentation quality assessment when no masks exist to compare with. We formulate the segmentation quality assessment process as an automated classification task to distinguish between good and poor-quality segmentation masks for more accurate gestational age estimation. We validate the performance of our proposed approach on two datasets we collect from two hospitals using different ultrasound machines. We compare different architectures, with our best-performing architecture achieving over 90% classification accuracy on distinguishing between good and poor-quality segmentation masks from an unseen dataset. Additionally, there was only a 1.45-day difference between the gestational age reported by doctors and estimated based on CRL measurements using well-segmented masks. On the other hand, this difference increased and reached up to 7.73 days when we calculated CRL from the poorly segmented masks. As a result, AI-based approaches can potentially aid fetal ultrasound segmentation quality assessment and might detect poor segmentation in real-time screening in the future.
comment: 13 pages, 3 figures, 3 tables
☆ PL-UNeXt: Per-stage Edge Detail and Line Feature Guided Segmentation for Power Line Detection
Power line detection is a critical inspection task for electricity companies and is also useful in avoiding drone obstacles. Accurately separating power lines from the surrounding area in the aerial image is still challenging due to the intricate background and low pixel ratio. In order to properly capture the guidance of the spatial edge detail prior and line features, we offer PL-UNeXt, a power line segmentation model with a booster training strategy. We design edge detail heads computing the loss in edge space to guide the lower-level detail learning and line feature heads generating auxiliary segmentation masks to supervise higher-level line feature learning. Benefited from this design, our model can reach 70.6 F1 score (+1.9%) on TTPLA and 68.41 mIoU (+5.2%) on VITL (without utilizing IR images), while preserving a real-time performance due to few inference parameters.
☆ Intermediate and Future Frame Prediction of Geostationary Satellite Imagery With Warp and Refine Network AAAI2022
Geostationary satellite imagery has applications in climate and weather forecasting, planning natural energy resources, and predicting extreme weather events. For precise and accurate prediction, higher spatial and temporal resolution of geostationary satellite imagery is important. Although recent geostationary satellite resolution has improved, the long-term analysis of climate applications is limited to using multiple satellites from the past to the present due to the different resolutions. To solve this problem, we proposed warp and refine network (WR-Net). WR-Net is divided into an optical flow warp component and a warp image refinement component. We used the TV-L1 algorithm instead of deep learning-based approaches to extract the optical flow warp component. The deep-learning-based model is trained on the human-centric view of the RGB channel and does not work on geostationary satellites, which is gray-scale one-channel imagery. The refinement network refines the warped image through a multi-temporal fusion layer. We evaluated WR-Net by interpolation of temporal resolution at 4 min intervals to 2 min intervals in large-scale GK2A geostationary meteorological satellite imagery. Furthermore, we applied WR-Net to the future frame prediction task and showed that the explicit use of optical flow can help future frame prediction.
comment: This paper has been accepted for the AAAI2022 Climate Change Workshop
☆ Imbalanced Open Set Domain Adaptation via Moving-threshold Estimation and Gradual Alignment
Multimedia applications are often associated with cross-domain knowledge transfer, where Unsupervised Domain Adaptation (UDA) can be used to reduce the domain shifts. Open Set Domain Adaptation (OSDA) aims to transfer knowledge from a well-labeled source domain to an unlabeled target domain under the assumption that the target domain contains unknown classes. Existing OSDA methods consistently lay stress on the covariate shift, ignoring the potential label shift problem. The performance of OSDA methods degrades drastically under intra-domain class imbalance and inter-domain label shift. However, little attention has been paid to this issue in the community. In this paper, the Imbalanced Open Set Domain Adaptation (IOSDA) is explored where the covariate shift, label shift and category mismatch exist simultaneously. To alleviate the negative effects raised by label shift in OSDA, we propose Open-set Moving-threshold Estimation and Gradual Alignment (OMEGA) - a novel architecture that improves existing OSDA methods on class-imbalanced data. Specifically, a novel unknown-aware target clustering scheme is proposed to form tight clusters in the target domain to reduce the negative effects of label shift and intra-domain class imbalance. Furthermore, moving-threshold estimation is designed to generate specific thresholds for each target sample rather than using one for all. Extensive experiments on IOSDA, OSDA and OPDA benchmarks demonstrate that our method could significantly outperform existing state-of-the-arts. Code and data are available at https://github.com/mendicant04/OMEGA.
comment: 11 pages, 5 figures, 7 tables
☆ Interpretable Visual Question Answering Referring to Outside Knowledge
We present a novel multimodal interpretable VQA model that can answer the question more accurately and generate diverse explanations. Although researchers have proposed several methods that can generate human-readable and fine-grained natural language sentences to explain a model's decision, these methods have focused solely on the information in the image. Ideally, the model should refer to various information inside and outside the image to correctly generate explanations, just as we use background knowledge daily. The proposed method incorporates information from outside knowledge and multiple image captions to increase the diversity of information available to the model. The contribution of this paper is to construct an interpretable visual question answering model using multimodal inputs to improve the rationality of generated results. Experimental results show that our model can outperform state-of-the-art methods regarding answer accuracy and explanation rationality.
comment: Under review
☆ SEMv2: Table Separation Line Detection Based on Conditional Convolution
Table structure recognition is an indispensable element for enabling machines to comprehend tables. Its primary purpose is to identify the internal structure of a table. Nevertheless, due to the complexity and diversity of their structure and style, it is highly challenging to parse the tabular data into a structured format that machines can comprehend. In this work, we adhere to the principle of the split-and-merge based methods and propose an accurate table structure recognizer, termed SEMv2 (SEM: Split, Embed and Merge). Unlike the previous works in the ``split'' stage, we aim to address the table separation line instance-level discrimination problem and introduce a table separation line detection strategy based on conditional convolution. Specifically, we design the ``split'' in a top-down manner that detects the table separation line instance first and then dynamically predicts the table separation line mask for each instance. The final table separation line shape can be accurately obtained by processing the table separation line mask in a row-wise/column-wise manner. To comprehensively evaluate the SEMv2, we also present a more challenging dataset for table structure recognition, dubbed iFLYTAB, which encompasses multiple style tables in various scenarios such as photos, scanned documents, etc. Extensive experiments on publicly available datasets (e.g. SciTSR, PubTabNet and iFLYTAB) demonstrate the efficacy of our proposed approach. The code and iFLYTAB dataset will be made publicly available upon acceptance of this paper.
☆ SGDViT: Saliency-Guided Dynamic Vision Transformer for UAV Tracking
Vision-based object tracking has boosted extensive autonomous applications for unmanned aerial vehicles (UAVs). However, the dynamic changes in flight maneuver and viewpoint encountered in UAV tracking pose significant difficulties, e.g. , aspect ratio change, and scale variation. The conventional cross-correlation operation, while commonly used, has limitations in effectively capturing perceptual similarity and incorporates extraneous background information. To mitigate these limitations, this work presents a novel saliency-guided dynamic vision Transformer (SGDViT) for UAV tracking. The proposed method designs a new task-specific object saliency mining network to refine the cross-correlation operation and effectively discriminate foreground and background information. Additionally, a saliency adaptation embedding operation dynamically generates tokens based on initial saliency, thereby reducing the computational complexity of the Transformer architecture. Finally, a lightweight saliency filtering Transformer further refines saliency information and increases the focus on appearance information. The efficacy and robustness of the proposed approach have been thoroughly assessed through experiments on three widely-used UAV tracking benchmarks and real-world scenarios, with results demonstrating its superiority. The source code and demo videos are available at https://github.com/vision4robotics/SGDViT.
☆ TSANET: Temporal and Scale Alignment for Unsupervised Video Object Segmentation
Unsupervised Video Object Segmentation (UVOS) refers to the challenging task of segmenting the prominent object in videos without manual guidance. In other words, the network detects the accurate region of the target object in a sequence of RGB frames without prior knowledge. In recent works, two approaches for UVOS have been discussed that can be divided into: appearance and appearance-motion based methods. Appearance based methods utilize the correlation information of inter-frames to capture target object that commonly appears in a sequence. However, these methods does not consider the motion of target object due to exploit the correlation information between randomly paired frames. Appearance-motion based methods, on the other hand, fuse the appearance features from RGB frames with the motion features from optical flow. Motion cue provides useful information since salient objects typically show distinctive motion in a sequence. However, these approaches have the limitation that the dependency on optical flow is dominant. In this paper, we propose a novel framework for UVOS that can address aforementioned limitations of two approaches in terms of both time and scale. Temporal Alignment Fusion aligns the saliency information of adjacent frames with the target frame to leverage the information of adjacent frames. Scale Alignment Decoder predicts the target object mask precisely by aggregating differently scaled feature maps via continuous mapping with implicit neural representation. We present experimental results on public benchmark datasets, DAVIS 2016 and FBMS, which demonstrate the effectiveness of our method. Furthermore, we outperform the state-of-the-art methods on DAVIS 2016.
☆ SANDFORMER: CNN and Transformer under Gated Fusion for Sand Dust Image Restoration ICASSP 2023
Although Convolutional Neural Networks (CNN) have made good progress in image restoration, the intrinsic equivalence and locality of convolutions still constrain further improvements in image quality. Recent vision transformer and self-attention have achieved promising results on various computer vision tasks. However, directly utilizing Transformer for image restoration is a challenging task. In this paper, we introduce an effective hybrid architecture for sand image restoration tasks, which leverages local features from CNN and long-range dependencies captured by transformer to improve the results further. We propose an efficient hybrid structure for sand dust image restoration to solve the feature inconsistency issue between Transformer and CNN. The framework complements each representation by modulating features from the CNN-based and Transformer-based branches rather than simply adding or concatenating features. Experiments demonstrate that SandFormer achieves significant performance improvements in synthetic and real dust scenes compared to previous sand image restoration methods.
comment: ICASSP 2023
☆ Dynamic Scenario Representation Learning for Motion Forecasting with Heterogeneous Graph Convolutional Recurrent Networks
Due to the complex and changing interactions in dynamic scenarios, motion forecasting is a challenging problem in autonomous driving. Most existing works exploit static road graphs to characterize scenarios and are limited in modeling evolving spatio-temporal dependencies in dynamic scenarios. In this paper, we resort to dynamic heterogeneous graphs to model the scenario. Various scenario components including vehicles (agents) and lanes, multi-type interactions, and their changes over time are jointly encoded. Furthermore, we design a novel heterogeneous graph convolutional recurrent network, aggregating diverse interaction information and capturing their evolution, to learn to exploit intrinsic spatio-temporal dependencies in dynamic graphs and obtain effective representations of dynamic scenarios. Finally, with a motion forecasting decoder, our model predicts realistic and multi-modal future trajectories of agents and outperforms state-of-the-art published works on several motion forecasting benchmarks.
☆ Sample Efficient Multimodal Semantic Augmentation for Incremental Summarization
In this work, we develop a prompting approach for incremental summarization of task videos. We develop a sample-efficient few-shot approach for extracting semantic concepts as an intermediate step. We leverage an existing model for extracting the concepts from the images and extend it to videos and introduce a clustering and querying approach for sample efficiency, motivated by the recent advances in perceiver-based architectures. Our work provides further evidence that an approach with richer input context with relevant entities and actions from the videos and using these as prompts could enhance the summaries generated by the model. We show the results on a relevant dataset and discuss possible directions for the work.
☆ ElC-OIS: Ellipsoidal Clustering for Open-World Instance Segmentation on LiDAR Data
Open-world Instance Segmentation (OIS) is a challenging task that aims to accurately segment every object instance appearing in the current observation, regardless of whether these instances have been labeled in the training set. This is important for safety-critical applications such as robust autonomous navigation. In this paper, we present a flexible and effective OIS framework for LiDAR point cloud that can accurately segment both known and unknown instances (i.e., seen and unseen instance categories during training). It first identifies points belonging to known classes and removes the background by leveraging close-set panoptic segmentation networks. Then, we propose a novel ellipsoidal clustering method that is more adapted to the characteristic of LiDAR scans and allows precise segmentation of unknown instances. Furthermore, a diffuse searching method is proposed to handle the common over-segmentation problem presented in the known instances. With the combination of these techniques, we are able to achieve accurate segmentation for both known and unknown instances. We evaluated our method on the SemanticKITTI open-world LiDAR instance segmentation dataset. The experimental results suggest that it outperforms current state-of-the-art methods, especially with a 10.0% improvement in association quality. The source code of our method will be publicly available at https://github.com/nubot-nudt/ElC-OIS.
☆ Semi-Supervised 2D Human Pose Estimation Driven by Position Inconsistency Pseudo Label Correction Module
In this paper, we delve into semi-supervised 2D human pose estimation. The previous method ignored two problems: (i) When conducting interactive training between large model and lightweight model, the pseudo label of lightweight model will be used to guide large models. (ii) The negative impact of noise pseudo labels on training. Moreover, the labels used for 2D human pose estimation are relatively complex: keypoint category and keypoint position. To solve the problems mentioned above, we propose a semi-supervised 2D human pose estimation framework driven by a position inconsistency pseudo label correction module (SSPCM). We introduce an additional auxiliary teacher and use the pseudo labels generated by the two teacher model in different periods to calculate the inconsistency score and remove outliers. Then, the two teacher models are updated through interactive training, and the student model is updated using the pseudo labels generated by two teachers. To further improve the performance of the student model, we use the semi-supervised Cut-Occlude based on pseudo keypoint perception to generate more hard and effective samples. In addition, we also proposed a new indoor overhead fisheye human keypoint dataset WEPDTOF-Pose. Extensive experiments demonstrate that our method outperforms the previous best semi-supervised 2D human pose estimation method. We will release the code and dataset at https://github.com/hlz0606/SSPCM.
☆ M-EBM: Towards Understanding the Manifolds of Energy-Based Models PAKDD 2023
Energy-based models (EBMs) exhibit a variety of desirable properties in predictive tasks, such as generality, simplicity and compositionality. However, training EBMs on high-dimensional datasets remains unstable and expensive. In this paper, we present a Manifold EBM (M-EBM) to boost the overall performance of unconditional EBM and Joint Energy-based Model (JEM). Despite its simplicity, M-EBM significantly improves unconditional EBMs in training stability and speed on a host of benchmark datasets, such as CIFAR10, CIFAR100, CelebA-HQ, and ImageNet 32x32. Once class labels are available, label-incorporated M-EBM (M-JEM) further surpasses M-EBM in image generation quality with an over 40% FID improvement, while enjoying improved accuracy. The code can be found at https://github.com/sndnyang/mebm.
comment: Accepted to PAKDD 2023
☆ Neural Vector Fields: Implicit Representation by Explicit Learning
Deep neural networks (DNNs) are widely applied for nowadays 3D surface reconstruction tasks and such methods can be further divided into two categories, which respectively warp templates explicitly by moving vertices or represent 3D surfaces implicitly as signed or unsigned distance functions. Taking advantage of both advanced explicit learning process and powerful representation ability of implicit functions, we propose a novel 3D representation method, Neural Vector Fields (NVF). It not only adopts the explicit learning process to manipulate meshes directly, but also leverages the implicit representation of unsigned distance functions (UDFs) to break the barriers in resolution and topology. Specifically, our method first predicts the displacements from queries towards the surface and models the shapes as \textit{Vector Fields}. Rather than relying on network differentiation to obtain direction fields as most existing UDF-based methods, the produced vector fields encode the distance and direction fields both and mitigate the ambiguity at "ridge" points, such that the calculation of direction fields is straightforward and differentiation-free. The differentiation-free characteristic enables us to further learn a shape codebook via Vector Quantization, which encodes the cross-object priors, accelerates the training procedure, and boosts model generalization on cross-category reconstruction. The extensive experiments on surface reconstruction benchmarks indicate that our method outperforms those state-of-the-art methods in different evaluation scenarios including watertight vs non-watertight shapes, category-specific vs category-agnostic reconstruction, category-unseen reconstruction, and cross-domain reconstruction. Our code will be publicly released.
☆ Privacy-preserving and Uncertainty-aware Federated Trajectory Prediction for Connected Autonomous Vehicles
Deep learning is the method of choice for trajectory prediction for autonomous vehicles. Unfortunately, its data-hungry nature implicitly requires the availability of sufficiently rich and high-quality centralized datasets, which easily leads to privacy leakage. Besides, uncertainty-awareness becomes increasingly important for safety-crucial cyber physical systems whose prediction module heavily relies on machine learning tools. In this paper, we relax the data collection requirement and enhance uncertainty-awareness by using Federated Learning on Connected Autonomous Vehicles with an uncertainty-aware global objective. We name our algorithm as FLTP. We further introduce ALFLTP which boosts FLTP via using active learning techniques in adaptatively selecting participating clients. We consider both negative log-likelihood (NLL) and aleatoric uncertainty (AU) as client selection metrics. Experiments on Argoverse dataset show that FLTP significantly outperforms the model trained on local data. In addition, ALFLTP-AU converges faster in training regression loss and performs better in terms of NLL, minADE and MR than FLTP in most rounds, and has more stable round-wise performance than ALFLTP-NLL.
☆ QuickSRNet: Plain Single-Image Super-Resolution Architecture for Faster Inference on Mobile Platforms
In this work, we present QuickSRNet, an efficient super-resolution architecture for real-time applications on mobile platforms. Super-resolution clarifies, sharpens, and upscales an image to higher resolution. Applications such as gaming and video playback along with the ever-improving display capabilities of TVs, smartphones, and VR headsets are driving the need for efficient upscaling solutions. While existing deep learning-based super-resolution approaches achieve impressive results in terms of visual quality, enabling real-time DL-based super-resolution on mobile devices with compute, thermal, and power constraints is challenging. To address these challenges, we propose QuickSRNet, a simple yet effective architecture that provides better accuracy-to-latency trade-offs than existing neural architectures for single-image super resolution. We present training tricks to speed up existing residual-based super-resolution architectures while maintaining robustness to quantization. Our proposed architecture produces 1080p outputs via 2x upscaling in 2.2 ms on a modern smartphone, making it ideal for high-fps real-time applications.
comment: 16 pages
☆ Corner Detection Based on Multi-directional Gabor Filters with Multi-scales
Gabor wavelet is an essential tool for image analysis and computer vision tasks. Local structure tensors with multiple scales are widely used in local feature extraction. Our research indicates that the current corner detection method based on Gabor wavelets can not effectively apply to complex scenes. In this work, the capability of the Gabor function to discriminate the intensity changes of step edges, L-shaped corners, Y-shaped or T-shaped corners, X-shaped corners, and star-shaped corners are investigated. The properties of Gabor wavelets to suppress affine image transformation are investigated and obtained. Many properties for edges and corners were discovered, which prompted us to propose a new corner extraction method. To fully use the structural information from the tuned Gabor filters, a novel multi-directional structure tensor is constructed for corner detection, and a multi-scale corner measurement function is proposed to remove false candidate corners. Furthermore, we compare the proposed method with twelve current state-of-the-art methods, which exhibit optimal performance and practical application to 3D reconstruction with good application potential.
☆ A Threefold Review on Deep Semantic Segmentation: Efficiency-oriented, Temporal and Depth-aware design
Semantic image and video segmentation stand among the most important tasks in computer vision nowadays, since they provide a complete and meaningful representation of the environment by means of a dense classification of the pixels in a given scene. Recently, Deep Learning, and more precisely Convolutional Neural Networks, have boosted semantic segmentation to a new level in terms of performance and generalization capabilities. However, designing Deep Semantic Segmentation models is a complex task, as it may involve application-dependent aspects. Particularly, when considering autonomous driving applications, the robustness-efficiency trade-off, as well as intrinsic limitations - computational/memory bounds and data-scarcity - and constraints - real-time inference - should be taken into consideration. In this respect, the use of additional data modalities, such as depth perception for reasoning on the geometry of a scene, and temporal cues from videos to explore redundancy and consistency, are promising directions yet not explored to their full potential in the literature. In this paper, we conduct a survey on the most relevant and recent advances in Deep Semantic Segmentation in the context of vision for autonomous vehicles, from three different perspectives: efficiency-oriented model development for real-time operation, RGB-Depth data integration (RGB-D semantic segmentation), and the use of temporal information from videos in temporally-aware models. Our main objective is to provide a comprehensive discussion on the main methods, advantages, limitations, results and challenges faced from each perspective, so that the reader can not only get started, but also be up to date in respect to recent advances in this exciting and challenging research field.
☆ Camera-Radar Perception for Autonomous Vehicles and ADAS: Concepts, Datasets and Metrics
One of the main paths towards the reduction of traffic accidents is the increase in vehicle safety through driver assistance systems or even systems with a complete level of autonomy. In these types of systems, tasks such as obstacle detection and segmentation, especially the Deep Learning-based ones, play a fundamental role in scene understanding for correct and safe navigation. Besides that, the wide variety of sensors in vehicles nowadays provides a rich set of alternatives for improvement in the robustness of perception in challenging situations, such as navigation under lighting and weather adverse conditions. Despite the current focus given to the subject, the literature lacks studies on radar-based and radar-camera fusion-based perception. Hence, this work aims to carry out a study on the current scenario of camera and radar-based perception for ADAS and autonomous vehicles. Concepts and characteristics related to both sensors, as well as to their fusion, are presented. Additionally, we give an overview of the Deep Learning-based detection and segmentation tasks, and the main datasets, metrics, challenges, and open questions in vehicle perception.
♻ ☆ Foresee What You Will Learn: Data Augmentation for Domain Generalization in Non-stationary Environment AAAI 2023
Existing domain generalization aims to learn a generalizable model to perform well even on unseen domains. For many real-world machine learning applications, the data distribution often shifts gradually along domain indices. For example, a self-driving car with a vision system drives from dawn to dusk, with the sky darkening gradually. Therefore, the system must be able to adapt to changes in ambient illumination and continue to drive safely on the road. In this paper, we formulate such problems as Evolving Domain Generalization, where a model aims to generalize well on a target domain by discovering and leveraging the evolving pattern of the environment. We then propose Directional Domain Augmentation (DDA), which simulates the unseen target features by mapping source data as augmentations through a domain transformer. Specifically, we formulate DDA as a bi-level optimization problem and solve it through a novel meta-learning approach in the representation space. We evaluate the proposed method on both synthetic datasets and realworld datasets, and empirical results show that our approach can outperform other existing methods.
comment: 12 pages, 6 figures, accepted by AAAI 2023
♻ ☆ Discovering Closed-Loop Failures of Vision-Based Controllers via Reachability Analysis
Machine learning driven image-based controllers allow robotic systems to take intelligent actions based on the visual feedback from their environment. Understanding when these controllers might lead to system safety violations is important for their integration in safety-critical applications and engineering corrective safety measures for the system. Existing methods leverage simulation-based testing (or falsification) to find the failures of vision-based controllers, i.e., the visual inputs that lead to closed-loop safety violations. However, these techniques do not scale well to the scenarios involving high-dimensional and complex visual inputs, such as RGB images. In this work, we cast the problem of finding closed-loop vision failures as a Hamilton-Jacobi (HJ) reachability problem. Our approach blends simulation-based analysis with HJ reachability methods to compute an approximation of the backward reachable tube (BRT) of the system, i.e., the set of unsafe states for the system under vision-based controllers. Utilizing the BRT, we can tractably and systematically find the system states and corresponding visual inputs that lead to closed-loop failures. These visual inputs can be subsequently analyzed to find the input characteristics that might have caused the failure. Besides its scalability to high-dimensional visual inputs, an explicit computation of BRT allows the proposed approach to capture non-trivial system failures that are difficult to expose via random simulations. We demonstrate our framework on two case studies involving an RGB image-based neural network controller for (a) autonomous indoor navigation, and (b) autonomous aircraft taxiing.
♻ ☆ Soft labelling for semantic segmentation: Bringing coherence to label down-sampling
In semantic segmentation, training data down-sampling is commonly performed due to limited resources, the need to adapt image size to the model input, or improve data augmentation. This down-sampling typically employs different strategies for the image data and the annotated labels. Such discrepancy leads to mismatches between the down-sampled color and label images. Hence, the training performance significantly decreases as the down-sampling factor increases. In this paper, we bring together the down-sampling strategies for the image data and the training labels. To that aim, we propose a novel framework for label down-sampling via soft-labeling that better conserves label information after down-sampling. Therefore, fully aligning soft-labels with image data to keep the distribution of the sampled pixels. This proposal also produces reliable annotations for under-represented semantic classes. Altogether, it allows training competitive models at lower resolutions. Experiments show that the proposal outperforms other down-sampling strategies. Moreover, state-of-the-art performance is achieved for reference benchmarks, but employing significantly less computational resources than foremost approaches. This proposal enables competitive research for semantic segmentation under resource constraints.
♻ ☆ Local Contrast and Global Contextual Information Make Infrared Small Object Salient Again
Infrared small object detection (ISOS) aims to segment small objects only covered with several pixels from clutter background in infrared images. It's of great challenge due to: 1) small objects lack of sufficient intensity, shape and texture information; 2) small objects are easily lost in the process where detection models, say deep neural networks, obtain high-level semantic features and image-level receptive fields through successive downsampling. This paper proposes a reliable detection model for ISOS, dubbed UCFNet, which can handle well the two issues. It builds upon central difference convolution (CDC) and fast Fourier convolution (FFC). On one hand, CDC can effectively guide the network to learn the contrast information between small objects and the background, as the contrast information is very essential in human visual system dealing with the ISOS task. On the other hand, FFC can gain image-level receptive fields and extract global information while preventing small objects from being overwhelmed.Experiments on several public datasets demonstrate that our method significantly outperforms the state-of-the-art ISOS models, and can provide useful guidelines for designing better ISOS deep models. Code are available at https://github.com/wcyjerry/BasicISOS.
♻ ☆ 3D-Aware Object Goal Navigation via Simultaneous Exploration and Identification CVPR 2023
Object goal navigation (ObjectNav) in unseen environments is a fundamental task for Embodied AI. Agents in existing works learn ObjectNav policies based on 2D maps, scene graphs, or image sequences. Considering this task happens in 3D space, a 3D-aware agent can advance its ObjectNav capability via learning from fine-grained spatial information. However, leveraging 3D scene representation can be prohibitively unpractical for policy learning in this floor-level task, due to low sample efficiency and expensive computational cost. In this work, we propose a framework for the challenging 3D-aware ObjectNav based on two straightforward sub-policies. The two sub-polices, namely corner-guided exploration policy and category-aware identification policy, simultaneously perform by utilizing online fused 3D points as observation. Through extensive experiments, we show that this framework can dramatically improve the performance in ObjectNav through learning from 3D scene representation. Our framework achieves the best performance among all modular-based methods on the Matterport3D and Gibson datasets, while requiring (up to 30x) less computational cost for training.
comment: To appear in CVPR 2023
♻ ☆ CGI-Stereo: Accurate and Real-Time Stereo Matching via Context and Geometry Interaction
In this paper, we propose CGI-Stereo, a novel neural network architecture that can concurrently achieve real-time performance, competitive accuracy, and strong generalization ability. The core of our CGI-Stereo is a Context and Geometry Fusion (CGF) block which adaptively fuses context and geometry information for more effective cost aggregation and meanwhile provides feedback to feature learning to guide more effective contextual feature extraction. The proposed CGF can be easily embedded into many existing stereo matching networks, such as PSMNet, GwcNet and ACVNet. The resulting networks show a significant improvement in accuracy. Specially, the model which incorporates our CGF with ACVNet ranks $1^{st}$ on the KITTI 2012 and 2015 leaderboards among all the published methods. We further propose an informative and concise cost volume, named Attention Feature Volume (AFV), which exploits a correlation volume as attention weights to filter a feature volume. Based on CGF and AFV, the proposed CGI-Stereo outperforms all other published real-time methods on KITTI benchmarks and shows better generalization ability than other real-time methods. Code is available at https://github.com/gangweiX/CGI-Stereo.
comment: 10 pages, 6 figures
♻ ☆ SeekNet: Improved Human Instance Segmentation and Tracking via Reinforcement Learning Based Optimized Robot Relocation
Amodal recognition is the ability of the system to detect occluded objects. Most SOTA Visual Recognition systems lack the ability to perform amodal recognition. Few studies have achieved amodal recognition through passive prediction or embodied recognition approaches. However, these approaches suffer from challenges in real-world applications, such as dynamic obstacles. We propose SeekNet, an improved optimization method for amodal recognition through embodied visual recognition. Additionally, we implement SeekNet for social robots, where there are multiple interactions with crowded pedestrians. We also demonstrate the benefits of our algorithm on occluded human detection and tracking over other baselines. Additionally, we set up a multi-robot environment with SeekNet to identify and track visual disease markers for airborne disease in crowded areas. We conduct our experiments in a simulated indoor environment and show that our method enhances the overall accuracy of the amodal recognition task and achieves the largest improvement in detection accuracy over time in comparison to the baseline approaches.
♻ ☆ SHIFT15M: Fashion-specific dataset for set-to-set matching with several distribution shifts
This paper addresses the problem of set-to-set matching, which involves matching two different sets of items based on some criteria, especially in the case of high-dimensional items like images. Although neural networks have been applied to solve this problem, most machine learning-based approaches assume that the training and test data follow the same distribution, which is not always true in real-world scenarios. To address this limitation, we introduce SHIFT15M, a dataset that can be used to evaluate set-to-set matching models when the distribution of data changes between training and testing. We conduct benchmark experiments that demonstrate the performance drop of naive methods due to distribution shift. Additionally, we provide software to handle the SHIFT15M dataset in a simple manner, with the URL for the software to be made available after publication of this manuscript. We believe proposed SHIFT15M dataset provide a valuable resource for evaluating set-to-set matching models under the distribution shift.
♻ ☆ Interactive Cartoonization with Controllable Perceptual Factors CVPR 2023
Cartoonization is a task that renders natural photos into cartoon styles. Previous deep cartoonization methods only have focused on end-to-end translation, which may hinder editability. Instead, we propose a novel solution with editing features of texture and color based on the cartoon creation process. To do that, we design a model architecture to have separate decoders, texture and color, to decouple these attributes. In the texture decoder, we propose a texture controller, which enables a user to control stroke style and abstraction to generate diverse cartoon textures. We also introduce an HSV color augmentation to induce the networks to generate diverse and controllable color translation. To the best of our knowledge, our work is the first deep approach to control the cartoonization at inference while showing profound quality improvement over to baselines.
comment: CVPR 2023
♻ ☆ A geometrically aware auto-encoder for multi-texture synthesis
We propose an auto-encoder architecture for multi-texture synthesis. The approach relies on both a compact encoder accounting for second order neural statistics and a generator incorporating adaptive periodic content. Images are embedded in a compact and geometrically consistent latent space, where the texture representation and its spatial organisation are disentangled. Texture synthesis and interpolation tasks can be performed directly from these latent codes. Our experiments demonstrate that our model outperforms state-of-the-art feed-forward methods in terms of visual quality and various texture related metrics.
♻ ☆ CaDM: Codec-aware Diffusion Modeling for Neural-enhanced Video Streaming
Recent years have witnessed the dramatic growth of Internet video traffic, where the video bitstreams are often compressed and delivered in low quality to fit the streamer's uplink bandwidth. To alleviate the quality degradation, it comes the rise of Neural-enhanced Video Streaming (NVS), which shows great prospects for recovering low-quality videos by mostly deploying neural super-resolution (SR) on the media server. Despite its benefit, we reveal that current mainstream works with SR enhancement have not achieved the desired rate-distortion trade-off between bitrate saving and quality restoration, due to: (1) overemphasizing the enhancement on the decoder side while omitting the co-design of encoder, (2) limited generative capacity to recover high-fidelity perceptual details, and (3) optimizing the compression-and-restoration pipeline from the resolution perspective solely, without considering color bit-depth. Aiming at overcoming these limitations, we are the first to conduct an encoder-decoder (i.e., codec) synergy by leveraging the inherent visual-generative property of diffusion models. Specifically, we present the Codec-aware Diffusion Modeling (CaDM), a novel NVS paradigm to significantly reduce streaming delivery bitrates while holding pretty higher restoration capacity over existing methods. First, CaDM improves the encoder's compression efficiency by simultaneously reducing resolution and color bit-depth of video frames. Second, CaDM empowers the decoder with high-quality enhancement by making the denoising diffusion restoration aware of encoder's resolution-color conditions. Evaluation on public cloud services with OpenMMLab benchmarks shows that CaDM effectively saves up to 5.12 - 21.44 times bitrates based on common video standards and achieves much better recovery quality (e.g., FID of 0.61) over state-of-the-art neural-enhancing methods.
♻ ☆ Continuous Sign Language Recognition with Correlation Network CVPR2023
Human body trajectories are a salient cue to identify actions in the video. Such body trajectories are mainly conveyed by hands and face across consecutive frames in sign language. However, current methods in continuous sign language recognition (CSLR) usually process frames independently, thus failing to capture cross-frame trajectories to effectively identify a sign. To handle this limitation, we propose correlation network (CorrNet) to explicitly capture and leverage body trajectories across frames to identify signs. In specific, a correlation module is first proposed to dynamically compute correlation maps between the current frame and adjacent frames to identify trajectories of all spatial patches. An identification module is then presented to dynamically emphasize the body trajectories within these correlation maps. As a result, the generated features are able to gain an overview of local temporal movements to identify a sign. Thanks to its special attention on body trajectories, CorrNet achieves new state-of-the-art accuracy on four large-scale datasets, i.e., PHOENIX14, PHOENIX14-T, CSL-Daily, and CSL. A comprehensive comparison with previous spatial-temporal reasoning methods verifies the effectiveness of CorrNet. Visualizations demonstrate the effects of CorrNet on emphasizing human body trajectories across adjacent frames.
comment: CVPR2023, code: https://github.com/hulianyuyy/CorrNet. arXiv admin note: text overlap with arXiv:2211.17081
♻ ☆ Self-supervised learning with rotation-invariant kernels
We introduce a regularization loss based on kernel mean embeddings with rotation-invariant kernels on the hypersphere (also known as dot-product kernels) for self-supervised learning of image representations. Besides being fully competitive with the state of the art, our method significantly reduces time and memory complexity for self-supervised training, making it implementable for very large embedding dimensions on existing devices and more easily adjustable than previous methods to settings with limited resources. Our work follows the major paradigm where the model learns to be invariant to some predefined image transformations (cropping, blurring, color jittering, etc.), while avoiding a degenerate solution by regularizing the embedding distribution. Our particular contribution is to propose a loss family promoting the embedding distribution to be close to the uniform distribution on the hypersphere, with respect to the maximum mean discrepancy pseudometric. We demonstrate that this family encompasses several regularizers of former methods, including uniformity-based and information-maximization methods, which are variants of our flexible regularization loss with different kernels. Beyond its practical consequences for state-of-the-art self-supervised learning with limited resources, the proposed generic regularization approach opens perspectives to leverage more widely the literature on kernel methods in order to improve self-supervised learning methods.
♻ ☆ Learning Imbalanced Data with Vision Transformers CVPR 2023
The real-world data tends to be heavily imbalanced and severely skew the data-driven deep neural networks, which makes Long-Tailed Recognition (LTR) a massive challenging task. Existing LTR methods seldom train Vision Transformers (ViTs) with Long-Tailed (LT) data, while the off-the-shelf pretrain weight of ViTs always leads to unfair comparisons. In this paper, we systematically investigate the ViTs' performance in LTR and propose LiVT to train ViTs from scratch only with LT data. With the observation that ViTs suffer more severe LTR problems, we conduct Masked Generative Pretraining (MGP) to learn generalized features. With ample and solid evidence, we show that MGP is more robust than supervised manners. In addition, Binary Cross Entropy (BCE) loss, which shows conspicuous performance with ViTs, encounters predicaments in LTR. We further propose the balanced BCE to ameliorate it with strong theoretical groundings. Specially, we derive the unbiased extension of Sigmoid and compensate extra logit margins to deploy it. Our Bal-BCE contributes to the quick convergence of ViTs in just a few epochs. Extensive experiments demonstrate that with MGP and Bal-BCE, LiVT successfully trains ViTs well without any additional data and outperforms comparable state-of-the-art methods significantly, e.g., our ViT-B achieves 81.0% Top-1 accuracy in iNaturalist 2018 without bells and whistles. Code is available at https://github.com/XuZhengzhuo/LiVT.
comment: Accepted to CVPR 2023, camera-ready version; Code: https://github.com/XuZhengzhuo/LiVT
♻ ☆ Improved Trajectory Reconstruction for Markerless Pose Estimation
Markerless pose estimation allows reconstructing human movement from multiple synchronized and calibrated views, and has the potential to make movement analysis easy and quick, including gait analysis. This could enable much more frequent and quantitative characterization of gait impairments, allowing better monitoring of outcomes and responses to interventions. However, the impact of different keypoint detectors and reconstruction algorithms on markerless pose estimation accuracy has not been thoroughly evaluated. We tested these algorithmic choices on data acquired from a multicamera system from a heterogeneous sample of 25 individuals seen in a rehabilitation hospital. We found that using a top-down keypoint detector and reconstructing trajectories with an implicit function enabled accurate, smooth and anatomically plausible trajectories, with a noise in the step width estimates compared to a GaitRite walkway of only 8mm.
♻ ☆ Video Question Answering Using CLIP-Guided Visual-Text Attention ICIP 2023
Cross-modal learning of video and text plays a key role in Video Question Answering (VideoQA). In this paper, we propose a visual-text attention mechanism to utilize the Contrastive Language-Image Pre-training (CLIP) trained on lots of general domain language-image pairs to guide the cross-modal learning for VideoQA. Specifically, we first extract video features using a TimeSformer and text features using a BERT from the target application domain, and utilize CLIP to extract a pair of visual-text features from the general-knowledge domain through the domain-specific learning. We then propose a Cross-domain Learning to extract the attention information between visual and linguistic features across the target domain and general domain. The set of CLIP-guided visual-text features are integrated to predict the answer. The proposed method is evaluated on MSVD-QA and MSRVTT-QA datasets, and outperforms state-of-the-art methods.
comment: Submitted to the 2023 IEEE International Conference on Image Processing (ICIP 2023)
♻ ☆ Text2Face: A Multi-Modal 3D Face Model
We present the first 3D morphable modelling approach, whereby 3D face shape can be directly and completely defined using a textual prompt. Building on work in multi-modal learning, we extend the FLAME head model to a common image-and-text latent space. This allows for direct 3D Morphable Model (3DMM) parameter generation and therefore shape manipulation from textual descriptions. Our method, Text2Face, has many applications; for example: generating police photofits where the input is already in natural language. It further enables multi-modal 3DMM image fitting to sketches and sculptures, as well as images.
comment: Fixed formatting and a typo
♻ ☆ TRR360D: A dataset for 360 degree rotated rectangular box table detection
To address the problem of scarcity and high annotation costs of rotated image table detection datasets, this paper proposes a method for building a rotated image table detection dataset. Based on the ICDAR2019MTD modern table detection dataset, we refer to the annotation format of the DOTA dataset to create the TRR360D rotated table detection dataset. The training set contains 600 rotated images and 977 annotated instances, and the test set contains 240 rotated images and 499 annotated instances. The AP50(T<90) evaluation metric is defined, and this dataset is available for future researchers to study rotated table detection algorithms and promote the development of table detection technology. The TRR360D rotated table detection dataset was created by constraining the starting point and annotation direction, and is publicly available at https://github.com/vansin/TRR360D.
♻ ☆ Grounding Language with Visual Affordances over Unstructured Data ICRA
Recent works have shown that Large Language Models (LLMs) can be applied to ground natural language to a wide variety of robot skills. However, in practice, learning multi-task, language-conditioned robotic skills typically requires large-scale data collection and frequent human intervention to reset the environment or help correcting the current policies. In this work, we propose a novel approach to efficiently learn general-purpose language-conditioned robot skills from unstructured, offline and reset-free data in the real world by exploiting a self-supervised visuo-lingual affordance model, which requires annotating as little as 1% of the total data with language. We evaluate our method in extensive experiments both in simulated and real-world robotic tasks, achieving state-of-the-art performance on the challenging CALVIN benchmark and learning over 25 distinct visuomotor manipulation tasks with a single policy in the real world. We find that when paired with LLMs to break down abstract natural language instructions into subgoals via few-shot prompting, our method is capable of completing long-horizon, multi-tier tasks in the real world, while requiring an order of magnitude less data than previous approaches. Code and videos are available at http://hulc2.cs.uni-freiburg.de
comment: Accepted at the 2023 IEEE International Conference on Robotics and Automation (ICRA). Project website: http://hulc2.cs.uni-freiburg.de
♻ ☆ Visual Language Maps for Robot Navigation ICRA
Grounding language to the visual observations of a navigating agent can be performed using off-the-shelf visual-language models pretrained on Internet-scale data (e.g., image captions). While this is useful for matching images to natural language descriptions of object goals, it remains disjoint from the process of mapping the environment, so that it lacks the spatial precision of classic geometric maps. To address this problem, we propose VLMaps, a spatial map representation that directly fuses pretrained visual-language features with a 3D reconstruction of the physical world. VLMaps can be autonomously built from video feed on robots using standard exploration approaches and enables natural language indexing of the map without additional labeled data. Specifically, when combined with large language models (LLMs), VLMaps can be used to (i) translate natural language commands into a sequence of open-vocabulary navigation goals (which, beyond prior work, can be spatial by construction, e.g., "in between the sofa and TV" or "three meters to the right of the chair") directly localized in the map, and (ii) can be shared among multiple robots with different embodiments to generate new obstacle maps on-the-fly (by using a list of obstacle categories). Extensive experiments carried out in simulated and real world environments show that VLMaps enable navigation according to more complex language instructions than existing methods. Videos are available at https://vlmaps.github.io.
comment: Accepted at the 2023 IEEE International Conference on Robotics and Automation (ICRA). Project page: https://vlmaps.github.io
♻ ☆ Extreme Masking for Learning Instance and Distributed Visual Representations
The paper presents a scalable approach for learning spatially distributed visual representations over individual tokens and a holistic instance representation simultaneously. We use self-attention blocks to represent spatially distributed tokens, followed by cross-attention blocks to aggregate the holistic image instance. The core of the approach is the use of extremely large token masking (75\%-90\%) as the data augmentation for supervision. Our model, named ExtreMA, follows the plain BYOL approach where the instance representation from the unmasked subset is trained to predict that from the intact input. Instead of encouraging invariance across inputs, the model is required to capture informative variations in an image. The paper makes three contributions: 1) It presents random masking as a strong and computationally efficient data augmentation for siamese representation learning. 2) With multiple sampling per instance, extreme masking greatly speeds up learning and improves performance with more data. 3) ExtreMA obtains stronger linear probing performance than masked modeling methods, and better transfer performance than prior contrastive models.
comment: Accepted in TMLR
♻ ☆ GLOW: Global Layout Aware Attacks on Object Detection ICCV
Adversarial attacks aim to perturb images such that a predictor outputs incorrect results. Due to the limited research in structured attacks, imposing consistency checks on natural multi-object scenes is a promising yet practical defense against conventional adversarial attacks. More desired attacks, to this end, should be able to fool defenses with such consistency checks. Therefore, we present the first approach GLOW that copes with various attack requests by generating global layout-aware adversarial attacks, in which both categorical and geometric layout constraints are explicitly established. Specifically, we focus on object detection task and given a victim image, GLOW first localizes victim objects according to target labels. And then it generates multiple attack plans, together with their context-consistency scores. Our proposed GLOW, on the one hand, is capable of handling various types of requests, including single or multiple victim objects, with or without specified victim objects. On the other hand, it produces a consistency score for each attack plan, reflecting the overall contextual consistency that both semantic category and global scene layout are considered. In experiment, we design multiple types of attack requests and validate our ideas on MS COCO and Pascal. Extensive experimental results demonstrate that we can achieve about 30$\%$ average relative improvement compared to state-of-the-art methods in conventional single object attack request; Moreover, our method outperforms SOTAs significantly on more generic attack requests by about 20$\%$ in average; Finally, our method produces superior performance under challenging zero-query black-box setting, or 20$\%$ better than SOTAs. Our code, model and attack requests would be made available.
comment: ICCV
♻ ☆ Image Inpainting via Iteratively Decoupled Probabilistic Modeling
Generative adversarial networks (GANs) have made great success in image inpainting yet still have difficulties tackling large missing regions. In contrast, iterative probabilistic algorithms, such as autoregressive and denoising diffusion models, have to be deployed with massive computing resources for decent effect. To achieve high-quality results with low computational cost, we present a novel pixel spread model (PSM) that iteratively employs decoupled probabilistic modeling, combining the optimization efficiency of GANs with the prediction tractability of probabilistic models. As a result, our model selectively spreads informative pixels throughout the image in a few iterations, largely enhancing the completion quality and efficiency. On multiple benchmarks, we achieve new state-of-the-art performance. Code is released at https://github.com/fenglinglwb/PSM.
comment: 20 pages
♻ ☆ JND-Based Perceptual Optimization For Learned Image Compression
Recently, learned image compression schemes have achieved remarkable improvements in image fidelity (e.g., PSNR and MS-SSIM) compared to conventional hybrid image coding ones due to their high-efficiency non-linear transform, end-to-end optimization frameworks, etc. However, few of them take the Just Noticeable Difference (JND) characteristic of the Human Visual System (HVS) into account and optimize learned image compression towards perceptual quality. To address this issue, a JND-based perceptual quality loss is proposed. Considering that the amounts of distortion in the compressed image at different training epochs under different Quantization Parameters (QPs) are different, we develop a distortion-aware adjustor. After combining them together, we can better assign the distortion in the compressed image with the guidance of JND to preserve the high perceptual quality. All these designs enable the proposed method to be flexibly applied to various learned image compression schemes with high scalability and plug-and-play advantages. Experimental results on the Kodak dataset demonstrate that the proposed method has led to better perceptual quality than the baseline model under the same bit rate.
comment: 5 pages, 5 figures, conference
♻ ☆ Invariant Feature Coding using Tensor Product Representation
In this study, a novel feature coding method that exploits invariance for transformations represented by a finite group of orthogonal matrices is proposed. We prove that the group-invariant feature vector contains sufficient discriminative information when learning a linear classifier using convex loss minimization. Based on this result, a novel feature model that explicitly consider group action is proposed for principal component analysis and k-means clustering, which are commonly used in most feature coding methods, and global feature functions. Although the global feature functions are in general complex nonlinear functions, the group action on this space can be easily calculated by constructing these functions as tensor-product representations of basic representations, resulting in an explicit form of invariant feature functions. The effectiveness of our method is demonstrated on several image datasets.
comment: 26 pages, 41 figures
♻ ☆ PixCUE: Joint Uncertainty Estimation and Image Reconstruction in MRI using Deep Pixel Classification
Deep learning (DL) models are capable of successfully exploiting latent representations in MR data and have become state-of-the-art for accelerated MRI reconstruction. However, undersampling the measurements in k-space as well as the over- or under-parameterized and non-transparent nature of DL make these models exposed to uncertainty. Consequently, uncertainty estimation has become a major issue in DL MRI reconstruction. To estimate uncertainty, Monte Carlo (MC) inference techniques have become a common practice where multiple reconstructions are utilized to compute the variance in reconstruction as a measurement of uncertainty. However, these methods demand high computational costs as they require multiple inferences through the DL model. To this end, we introduce a method to estimate uncertainty during MRI reconstruction using a pixel classification framework. The proposed method, PixCUE (stands for Pixel Classification Uncertainty Estimation) produces the reconstructed image along with an uncertainty map during a single forward pass through the DL model. We demonstrate that this approach generates uncertainty maps that highly correlate with the reconstruction errors with respect to various MR imaging sequences and under numerous adversarial conditions. We also show that the estimated uncertainties are correlated to that of the conventional MC method. We further provide an empirical relationship between the uncertainty estimations using PixCUE and well-established reconstruction metrics such as NMSE, PSNR, and SSIM. We conclude that PixCUE is capable of reliably estimating the uncertainty in MRI reconstruction with a minimum additional computational cost.
comment: 19 pages, 7 figures, 1 table
♻ ☆ CleanCLIP: Mitigating Data Poisoning Attacks in Multimodal Contrastive Learning
Multimodal contrastive pretraining has been used to train multimodal representation models, such as CLIP, on large amounts of paired image-text data. However, previous studies have revealed that such models are vulnerable to backdoor attacks. Specifically, when trained on backdoored examples, CLIP learns spurious correlations between the embedded backdoor trigger and the target label, aligning their representations in the joint embedding space. Injecting even a small number of poisoned examples, such as 75 examples in 3 million pretraining data, can significantly manipulate the model's behavior, making it difficult to detect or unlearn such correlations. To address this issue, we propose CleanCLIP, a finetuning framework that weakens the learned spurious associations introduced by backdoor attacks by independently re-aligning the representations for individual modalities. We demonstrate that unsupervised finetuning using a combination of multimodal contrastive and unimodal self-supervised objectives for individual modalities can significantly reduce the impact of the backdoor attack. Additionally, we show that supervised finetuning on task-specific labeled image data removes the backdoor trigger from the CLIP vision encoder. We show empirically that CleanCLIP maintains model performance on benign examples while erasing a range of backdoor attacks on multimodal contrastive learning.
comment: 20 pages, 7 figures, 8 tables
♻ ☆ Relationship Quantification of Image Degradations
In this paper, we study two challenging but less-touched problems in image restoration, namely, i) how to quantify the relationship between different image degradations and ii) how to improve the performance on a specific degradation using the quantified relationship. To tackle the first challenge, Degradation Relationship Index (DRI) is proposed to measure the degradation relationship, which is defined as the mean drop rate difference in the validation loss between two models, i.e., one is trained using the anchor degradation only and another is trained based on both the anchor and the auxiliary degradations. Through quantifying the relationship between different degradations using DRI, we empirically observe that i) the degradation combination proportion is crucial to the image restoration performance. In other words, the combinations with only appropriate degradation proportions could improve the performance of the anchor restoration; ii) a positive DRI always predicts the performance improvement of image restoration. Based on the observations, we propose an adaptive Degradation Proportion Determination strategy (DPD) which could improve the performance on the anchor degradation with the assist of another auxiliary degradation. Extensive experimental results verify the effective of our method by taking haze as the anchor degradation and noise, rain streak, and snow as the auxiliary degradations. The code will be released after acceptance.
♻ ☆ VoP: Text-Video Co-operative Prompt Tuning for Cross-Modal Retrieval CVPR 2023
Many recent studies leverage the pre-trained CLIP for text-video cross-modal retrieval by tuning the backbone with additional heavy modules, which not only brings huge computational burdens with much more parameters, but also leads to the knowledge forgetting from upstream models.In this work, we propose the VoP: Text-Video Co-operative Prompt Tuning for efficient tuning on the text-video retrieval task. The proposed VoP is an end-to-end framework with both video & text prompts introducing, which can be regarded as a powerful baseline with only 0.1% trainable parameters. Further, based on the spatio-temporal characteristics of videos, we develop three novel video prompt mechanisms to improve the performance with different scales of trainable parameters. The basic idea of the VoP enhancement is to model the frame position, frame context, and layer function with specific trainable prompts, respectively. Extensive experiments show that compared to full fine-tuning, the enhanced VoP achieves a 1.4% average R@1 gain across five text-video retrieval benchmarks with 6x less parameter overhead. The code will be available at https://github.com/bighuang624/VoP.
comment: Accepted by CVPR 2023
♻ ☆ DistilPose: Tokenized Pose Regression with Heatmap Distillation CVPR2023
In the field of human pose estimation, regression-based methods have been dominated in terms of speed, while heatmap-based methods are far ahead in terms of performance. How to take advantage of both schemes remains a challenging problem. In this paper, we propose a novel human pose estimation framework termed DistilPose, which bridges the gaps between heatmap-based and regression-based methods. Specifically, DistilPose maximizes the transfer of knowledge from the teacher model (heatmap-based) to the student model (regression-based) through Token-distilling Encoder (TDE) and Simulated Heatmaps. TDE aligns the feature spaces of heatmap-based and regression-based models by introducing tokenization, while Simulated Heatmaps transfer explicit guidance (distribution and confidence) from teacher heatmaps into student models. Extensive experiments show that the proposed DistilPose can significantly improve the performance of the regression-based models while maintaining efficiency. Specifically, on the MSCOCO validation dataset, DistilPose-S obtains 71.6% mAP with 5.36M parameter, 2.38 GFLOPs and 40.2 FPS, which saves 12.95x, 7.16x computational cost and is 4.9x faster than its teacher model with only 0.9 points performance drop. Furthermore, DistilPose-L obtains 74.4% mAP on MSCOCO validation dataset, achieving a new state-of-the-art among predominant regression-based models.
comment: accepted by CVPR2023
♻ ☆ IRNet: Iterative Refinement Network for Noisy Partial Label Learning
Partial label learning (PLL) is a typical weakly supervised learning, where each sample is associated with a set of candidate labels. The basic assumption of PLL is that the ground-truth label must reside in the candidate set. However, this assumption may not be satisfied due to the unprofessional judgment of the annotators, thus limiting the practical application of PLL. In this paper, we relax this assumption and focus on a more general problem, noisy PLL, where the ground-truth label may not exist in the candidate set. To address this challenging problem, we propose a novel framework called "Iterative Refinement Network (IRNet)". It aims to purify the noisy samples by two key modules, i.e., noisy sample detection and label correction. Ideally, we can convert noisy PLL into traditional PLL if all noisy samples are corrected. To guarantee the performance of these modules, we start with warm-up training and exploit data augmentation to reduce prediction errors. Through theoretical analysis, we prove that IRNet is able to reduce the noise level of the dataset and eventually approximate the Bayes optimal classifier. Experimental results on multiple benchmark datasets demonstrate the effectiveness of our method. IRNet is superior to existing state-of-the-art approaches on noisy PLL.
♻ ☆ Local Label Point Correction for Edge Detection of Overlapping Cervical Cells
Accurate labeling is essential for supervised deep learning methods. However, it is almost impossible to accurately and manually annotate thousands of images, which results in many labeling errors for most datasets. We proposes a local label point correction (LLPC) method to improve annotation quality for edge detection and image segmentation tasks. Our algorithm contains three steps: gradient-guided point correction, point interpolation and local point smoothing. We correct the labels of object contours by moving the annotated points to the pixel gradient peaks. This can improve the edge localization accuracy, but it also causes unsmooth contours due to the interference of image noise. Therefore, we design a point smoothing method based on local linear fitting to smooth the corrected edge. To verify the effectiveness of our LLPC, we construct a largest overlapping cervical cell edge detection dataset (CCEDD) with higher precision label corrected by our label correction method. Our LLPC only needs to set three parameters, but yields 30-40$\%$ average precision improvement on multiple networks. The qualitative and quantitative experimental results show that our LLPC can improve the quality of manual labels and the accuracy of overlapping cell edge detection. We hope that our study will give a strong boost to the development of the label correction for edge detection and image segmentation. We will release the dataset and code at https://github.com/nachifur/LLPC.
comment: Published on Frontiers in Neuroinformatics. Official paper: https://www.frontiersin.org/articles/10.3389/fninf.2022.895290/full. Code and dataset: https://github.com/nachifur/LLPC
♻ ☆ DexGraspNet: A Large-Scale Robotic Dexterous Grasp Dataset for General Objects Based on Simulation
Robotic dexterous grasping is the first step to enable human-like dexterous object manipulation and thus a crucial robotic technology. However, dexterous grasping is much more under-explored than object grasping with parallel grippers, partially due to the lack of a large-scale dataset. In this work, we present a large-scale robotic dexterous grasp dataset, DexGraspNet, generated by our proposed highly efficient synthesis method that can be generally applied to any dexterous hand. Our method leverages a deeply accelerated differentiable force closure estimator and thus can efficiently and robustly synthesize stable and diverse grasps on a large scale. We choose ShadowHand and generate 1.32 million grasps for 5355 objects, covering more than 133 object categories and containing more than 200 diverse grasps for each object instance, with all grasps having been validated by the Isaac Gym simulator. Compared to the previous dataset from Liu et al. generated by GraspIt!, our dataset has not only more objects and grasps, but also higher diversity and quality. Via performing cross-dataset experiments, we show that training several algorithms of dexterous grasp synthesis on our dataset significantly outperforms training on the previous one. To access our data and code, including code for human and Allegro grasp synthesis, please visit our project page: https://pku-epic.github.io/DexGraspNet/.
♻ ☆ Focus On Details: Online Multi-object Tracking with Diverse Fine-grained Representation CVPR2023
Discriminative representation is essential to keep a unique identifier for each target in Multiple object tracking (MOT). Some recent MOT methods extract features of the bounding box region or the center point as identity embeddings. However, when targets are occluded, these coarse-grained global representations become unreliable. To this end, we propose exploring diverse fine-grained representation, which describes appearance comprehensively from global and local perspectives. This fine-grained representation requires high feature resolution and precise semantic information. To effectively alleviate the semantic misalignment caused by indiscriminate contextual information aggregation, Flow Alignment FPN (FAFPN) is proposed for multi-scale feature alignment aggregation. It generates semantic flow among feature maps from different resolutions to transform their pixel positions. Furthermore, we present a Multi-head Part Mask Generator (MPMG) to extract fine-grained representation based on the aligned feature maps. Multiple parallel branches of MPMG allow it to focus on different parts of targets to generate local masks without label supervision. The diverse details in target masks facilitate fine-grained representation. Eventually, benefiting from a Shuffle-Group Sampling (SGS) training strategy with positive and negative samples balanced, we achieve state-of-the-art performance on MOT17 and MOT20 test sets. Even on DanceTrack, where the appearance of targets is extremely similar, our method significantly outperforms ByteTrack by 5.0% on HOTA and 5.6% on IDF1. Extensive experiments have proved that diverse fine-grained representation makes Re-ID great again in MOT.
comment: Accepted by CVPR2023
♻ ☆ Bort: Towards Explainable Neural Networks with Bounded Orthogonal Constraint ICLR 2023
Deep learning has revolutionized human society, yet the black-box nature of deep neural networks hinders further application to reliability-demanded industries. In the attempt to unpack them, many works observe or impact internal variables to improve the comprehensibility and invertibility of the black-box models. However, existing methods rely on intuitive assumptions and lack mathematical guarantees. To bridge this gap, we introduce Bort, an optimizer for improving model explainability with boundedness and orthogonality constraints on model parameters, derived from the sufficient conditions of model comprehensibility and invertibility. We perform reconstruction and backtracking on the model representations optimized by Bort and observe a clear improvement in model explainability. Based on Bort, we are able to synthesize explainable adversarial samples without additional parameters and training. Surprisingly, we find Bort constantly improves the classification accuracy of various architectures including ResNet and DeiT on MNIST, CIFAR-10, and ImageNet. Code: https://github.com/zbr17/Bort.
comment: ICLR 2023 accepted
♻ ☆ Research on the quantity and brightness evolution characteristics of Photospheric Bright Points groups
Context. Photospheric bright points (BPs), as the smallest magnetic element of the photosphere and the footpoint tracer of the magnetic flux tube, are of great significance to the study of BPs. Compared with the study of the characteristics and evolution of a few specific BPs, the study of BPs groups can provide us with a better understanding of the characteristics and overall activities of BPs groups. Aims. We aim to find out the evolution characteristics of the brightness and number of BPs groups at different brightness levels, and how these characteristics differ between quiet and active regions. Methods. We propose a hybrid BPs detection model (HBD Model) combining traditional technology and neural network. The Model is used to detect and calculate the BPs brightness characteristics of each frame of continuous high resolution image sequences of active and quiet regions in TiO-band of a pair of BBSO. Using machine learning clustering method, the PBs of each frame was divided into four levels groups (level1-level4) according to the brightness from low to high. Finally, Fourier transform and inverse Fourier transform are used to analyze the evolution of BPs brightness and quantity in these four levels groups. Results. The activities of BPs groups are not random and disorderly. In different levels of brightness, their quantity and brightness evolution show complex changes. Among the four levels of brightness, BPs in the active region were more active and intense than those in the quiet region. However, the quantity and brightness evolution of BPs groups in the quiet region showed the characteristics of large periodic changes and small periodic changes in the medium and high brightness levels (level3 and level4). The brightness evolution of PBs group in the quiet region has obvious periodic changes, but the active region is in a completely random and violent fluctuation state.
comment: The paper was edited using the Latex template from Astronomy & Astrophysics, but this paper has never been published publicly in Astronomy & Astrophysics. This mistake is very misleading. At the same time, there are some errors in the description of the experimental data in the original paper, and the experimental content is also insufficient
♻ ☆ Cross-Modal Learning with 3D Deformable Attention for Action Recognition
An important challenge in vision-based action recognition is the embedding of spatiotemporal features with two or more heterogeneous modalities into a single feature. In this study, we propose a new 3D deformable transformer for action recognition with adaptive spatiotemporal receptive fields and a cross-modal learning scheme. The 3D deformable transformer consists of three attention modules: 3D deformability, local joint stride, and temporal stride attention. The two cross-modal tokens are input into the 3D deformable attention module to create a cross-attention token with a reflected spatiotemporal correlation. Local joint stride attention is applied to spatially combine attention and pose tokens. Temporal stride attention temporally reduces the number of input tokens in the attention module and supports temporal expression learning without the simultaneous use of all tokens. The deformable transformer iterates L times and combines the last cross-modal token for classification. The proposed 3D deformable transformer was tested on the NTU60, NTU120, FineGYM, and PennAction datasets, and showed results better than or similar to pre-trained state-of-the-art methods even without a pre-training process. In addition, by visualizing important joints and correlations during action recognition through spatial joint and temporal stride attention, the possibility of achieving an explainable potential for action recognition is presented.
comment: 10 pages, 8 figures
Information Retrieval
☆ FastFill: Efficient Compatible Model Update
In many retrieval systems the original high dimensional data (e.g., images) is mapped to a lower dimensional feature through a learned embedding model. The task of retrieving the most similar data from a gallery set to a given query data is performed through a similarity comparison on features. When the embedding model is updated, it might produce features that are not comparable/compatible with features already in the gallery computed with the old model. Subsequently, all features in the gallery need to be re-computed using the new embedding model -- a computationally expensive process called backfilling. Recently, compatible representation learning methods have been proposed to avoid backfilling. Despite their relative success, there is an inherent trade-off between the new model performance and its compatibility with the old model. In this work, we introduce FastFill: a compatible model update process using feature alignment and policy based partial backfilling to promptly elevate retrieval performance. We show that previous backfilling strategies suffer from decreased performance and demonstrate the importance of both the training objective and the ordering in online partial backfilling. We propose a new training method for feature alignment between old and new embedding models using uncertainty estimation. Compared to previous works, we obtain significantly improved backfilling results on a variety of datasets: mAP on ImageNet (+4.4\%), Places-365 (+2.7\%), and VGG-Face2 (+1.3\%). Further, we demonstrate that when updating a biased model with FastFill, the minority subgroup accuracy gap promptly vanishes with a small fraction of partial backfilling.
comment: To appear in The Eleventh International Conference on Learning Representations
☆ A Prompt Log Analysis of Text-to-Image Generation Systems
Recent developments in diffusion models have unleashed the astonishing capabilities of text-to-image generation systems to synthesize high-quality images that are faithful to a given reference text, known as a "prompt." These systems, once released to the public, have immediately received tons of attention from researchers, creators, and common users. Despite the plenty of efforts to improve the underneath generative models, there is limited work on understanding the information needs of the real users of these systems, e.g., by investigating the prompts the users input at scale. In this paper, we take the initiative to conduct a comprehensive analysis of large-scale prompt logs collected from multiple text-to-image generation systems. Our work is analogous to analyzing the query log of Web search engines, a line of work that has made critical contributions to the glory of the Web search industry and research. We analyze over two million user-input prompts submitted to three popular text-to-image systems at scale. Compared to Web search queries, text-to-image prompts are significantly longer, often organized into unique structures, and present different categories of information needs. Users tend to make more edits within creation sessions, showing remarkable exploratory patterns. Our findings provide concrete implications on how to improve text-to-image generation systems for creation purposes.
☆ Kernel-CF: Collaborative filtering done right with social network analysis and kernel smoothing
Collaborative filtering is the simplest but oldest machine learning algorithm in the field of recommender systems. In spite of its long history, it remains a discussion topic in research venues. Usually people use users/items whose similarity scores with the target customer greater than 0 to compute the algorithms. However, this might not be the optimal solution after careful scrutiny. In this paper, we transform the recommender system input data into a 2-D social network, and apply kernel smoothing to compute preferences for unknown values in the user item rating matrix. We unifies the theoretical framework of recommender system and non-parametric statistics and provides an algorithmic procedure with optimal parameter selection method to achieve the goal.
☆ Class Cardinality Comparison as a Fermi Problem
Questions on class cardinality comparisons are quite tricky to answer and come with its own challenges. They require some kind of reasoning since web documents and knowledge bases, indispensable sources of information, rarely store direct answers to questions, such as, ``Are there more astronauts or Physics Nobel Laureates?'' We tackle questions on class cardinality comparison by tapping into three sources for absolute cardinalities as well as the cardinalities of orthogonal subgroups of the classes. We propose novel techniques for aggregating signals with partial coverage for more reliable estimates and evaluate them on a dataset of 4005 class pairs, achieving an accuracy of 83.7%.
comment: Accepted to the Web Conference 2023
☆ NASTyLinker: NIL-Aware Scalable Transformer-based Entity Linker ESWC'23
Entity Linking (EL) is the task of detecting mentions of entities in text and disambiguating them to a reference knowledge base. Most prevalent EL approaches assume that the reference knowledge base is complete. In practice, however, it is necessary to deal with the case of linking to an entity that is not contained in the knowledge base (NIL entity). Recent works have shown that, instead of focusing only on affinities between mentions and entities, considering inter-mention affinities can be used to represent NIL entities by producing clusters of mentions. At the same time, inter-mention affinities can help to substantially improve linking performance for known entities. With NASTyLinker, we introduce an EL approach that is aware of NIL-entities and produces corresponding mention clusters while maintaining high linking performance for known entities. The approach clusters mentions and entities based on dense representations from Transformers and resolves conflicts (if more than one entity is assigned to a cluster) by computing transitive mention-entity affinities. We show the effectiveness and scalability of NASTyLinker on NILK, a dataset that is explicitly constructed to evaluate EL with respect to NIL-entities. Further, we apply the presented approach to an actual EL task, namely to knowledge graph population by linking entities in Wikipedia listings, and provide an analysis of the outcome.
comment: Preprint of a paper in the research track of the 20th Extended Semantic Web Conference (ESWC'23)
☆ Achievable Rates and Low-Complexity Encoding of Posterior Matching for the BSC
Horstein, Burnashev, Shayevitz and Feder, Naghshvar et al. and others have studied sequential transmission of a K-bit message over the binary symmetric channel (BSC) with full, noiseless feedback using posterior matching. Yang et al. provide an improved lower bound on the achievable rate using martingale analysis that relies on the small-enough difference (SED) partitioning introduced by Naghshvar et al. SED requires a relatively complex encoder and decoder. To reduce complexity, this paper replaces SED with relaxed constraints that admit the small enough absolute difference (SEAD) partitioning rule. The main analytical results show that achievable-rate bounds higher than those found by Yang et al. are possible even under the new constraints, which are less restrictive than SED. The new analysis does not use martingale theory for the confirmation phase and applies a surrogate channel technique to tighten the results. An initial systematic transmission further increases the achievable rate bound. The simplified encoder associated with SEAD has a complexity below order O(K^2) and allows simulations for message sizes of at least 1000 bits. For example, simulations achieve 99% of of the channel's 0.50-bit capacity with an average block size of 200 bits for a target codeword error rate of 10^(-3).
comment: This paper consists of 26 pages and contains 6 figures. An earlier version of the algorithm included in this paper was published at the 2020 IEEE International Symposium on Information Theory (ISIT), (DOI: 10.1109/ISIT44484.2020.9174232)
☆ Unbiased Learning to Rank with Biased Continuous Feedback
It is a well-known challenge to learn an unbiased ranker with biased feedback. Unbiased learning-to-rank(LTR) algorithms, which are verified to model the relative relevance accurately based on noisy feedback, are appealing candidates and have already been applied in many applications with single categorical labels, such as user click signals. Nevertheless, the existing unbiased LTR methods cannot properly handle continuous feedback, which are essential for many industrial applications, such as content recommender systems. To provide personalized high-quality recommendation results, recommender systems need model both categorical and continuous biased feedback, such as click and dwell time. Accordingly, we design a novel unbiased LTR algorithm to tackle the challenges, which innovatively models position bias in the pairwise fashion and introduces the pairwise trust bias to separate the position bias, trust bias, and user relevance explicitly and can work for both continuous and categorical feedback. Experiment results on public benchmark datasets and internal live traffic of a large-scale recommender system at Tencent News show superior results for continuous labels and also competitive performance for categorical labels of the proposed method.
comment: 10 pages. arXiv admin note: substantial text overlap with arXiv:2111.12929
♻ ☆ GLCC: A General Framework for Graph-Level Clustering AAAI
This paper studies the problem of graph-level clustering, which is a novel yet challenging task. This problem is critical in a variety of real-world applications such as protein clustering and genome analysis in bioinformatics. Recent years have witnessed the success of deep clustering coupled with graph neural networks (GNNs). However, existing methods focus on clustering among nodes given a single graph, while exploring clustering on multiple graphs is still under-explored. In this paper, we propose a general graph-level clustering framework named Graph-Level Contrastive Clustering (GLCC) given multiple graphs. Specifically, GLCC first constructs an adaptive affinity graph to explore instance- and cluster-level contrastive learning (CL). Instance-level CL leverages graph Laplacian based contrastive loss to learn clustering-friendly representations while cluster-level CL captures discriminative cluster representations incorporating neighbor information of each sample. Moreover, we utilize neighbor-aware pseudo-labels to reward the optimization of representation learning. The two steps can be alternatively trained to collaborate and benefit each other. Experiments on a range of well-known datasets demonstrate the superiority of our proposed GLCC over competitive baselines.
comment: Accepted by Proceedings of the AAAI Conference on Artificial Intelligence (AAAI 2023)
♻ ☆ RETEXO: Scalable Neural Network Training over Distributed Graphs
Graph neural networks offer a promising approach to supervised learning over graph data. Graph data, especially when it is privacy-sensitive or too large to train on centrally, is often stored partitioned across disparate processing units (clients) which want to minimize the communication costs during collaborative training. The fully-distributed setup takes such partitioning to its extreme, wherein features of only a single node and its adjacent edges are kept locally with one client processor. Existing GNNs are not architected for training in such setups and incur prohibitive costs therein. We propose RETEXO, a novel transformation of existing GNNs that improves the communication efficiency during training in the fully-distributed setup. We experimentally confirm that RETEXO offers up to 6 orders of magnitude better communication efficiency even when training shallow GNNs, with a minimal trade-off in accuracy for supervised node classification tasks.
♻ ☆ Line Graph Contrastive Learning for Link Prediction
Link prediction tasks focus on predicting possible future connections. Most existing researches measure the likelihood of links by different similarity scores on node pairs and predict links between nodes. However, the similarity-based approaches have some challenges in information loss on nodes and generalization ability on similarity indexes. To address the above issues, we propose a Line Graph Contrastive Learning(LGCL) method to obtain rich information with multiple perspectives. LGCL obtains a subgraph view by h-hop subgraph sampling with target node pairs. After transforming the sampled subgraph into a line graph, the link prediction task is converted into a node classification task, which graph convolution progress can learn edge embeddings from graphs more effectively. Then we design a novel cross-scale contrastive learning framework on the line graph and the subgraph to maximize the mutual information of them, so that fuses the structure and feature information. The experimental results demonstrate that the proposed LGCL outperforms the state-of-the-art methods and has better performance on generalization and robustness.
comment: 37 pages
♻ ☆ Automatic Context Pattern Generation for Entity Set Expansion
Entity Set Expansion (ESE) is a valuable task that aims to find entities of the target semantic class described by given seed entities. Various Natural Language Processing (NLP) and Information Retrieval (IR) downstream applications have benefited from ESE due to its ability to discover knowledge. Although existing corpus-based ESE methods have achieved great progress, they still rely on corpora with high-quality entity information annotated, because most of them need to obtain the context patterns through the position of the entity in a sentence. Therefore, the quality of the given corpora and their entity annotation has become the bottleneck that limits the performance of such methods. To overcome this dilemma and make the ESE models free from the dependence on entity annotation, our work aims to explore a new ESE paradigm, namely corpus-independent ESE. Specifically, we devise a context pattern generation module that utilizes autoregressive language models (e.g., GPT-2) to automatically generate high-quality context patterns for entities. In addition, we propose the GAPA, a novel ESE framework that leverages the aforementioned GenerAted PAtterns to expand target entities. Extensive experiments and detailed analyses on three widely used datasets demonstrate the effectiveness of our method. All the codes of our experiments are available at https://github.com/geekjuruo/GAPA.
comment: This work has been submitted to the IEEE for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessible
Machine Learning
☆ Automatic Debiased Learning from Positive, Unlabeled, and Exposure Data
We address the issue of binary classification from positive and unlabeled data (PU classification) with a selection bias in the positive data. During the observation process, (i) a sample is exposed to a user, (ii) the user then returns the label for the exposed sample, and (iii) we however can only observe the positive samples. Therefore, the positive labels that we observe are a combination of both the exposure and the labeling, which creates a selection bias problem for the observed positive samples. This scenario represents a conceptual framework for many practical applications, such as recommender systems, which we refer to as ``learning from positive, unlabeled, and exposure data'' (PUE classification). To tackle this problem, we initially assume access to data with exposure labels. Then, we propose a method to identify the function of interest using a strong ignorability assumption and develop an ``Automatic Debiased PUE'' (ADPUE) learning method. This algorithm directly debiases the selection bias without requiring intermediate estimates, such as the propensity score, which is necessary for other learning methods. Through experiments, we demonstrate that our approach outperforms traditional PU learning methods on various semi-synthetic datasets.
☆ Ewald-based Long-Range Message Passing for Molecular Graphs
Neural architectures that learn potential energy surfaces from molecular data have undergone fast improvement in recent years. A key driver of this success is the Message Passing Neural Network (MPNN) paradigm. Its favorable scaling with system size partly relies upon a spatial distance limit on messages. While this focus on locality is a useful inductive bias, it also impedes the learning of long-range interactions such as electrostatics and van der Waals forces. To address this drawback, we propose Ewald message passing: a nonlocal Fourier space scheme which limits interactions via a cutoff on frequency instead of distance, and is theoretically well-founded in the Ewald summation method. It can serve as an augmentation on top of existing MPNN architectures as it is computationally cheap and agnostic to other architectural details. We test the approach with four baseline models and two datasets containing diverse periodic (OC20) and aperiodic structures (OE62). We observe robust improvements in energy mean absolute errors across all models and datasets, averaging 10% on OC20 and 16% on OE62. Our analysis shows an outsize impact of these improvements on structures with high long-range contributions to the ground truth energy.
☆ Enabling Non-Linear Quantum Operations through Variational Quantum Splines
The postulates of quantum mechanics impose only unitary transformations on quantum states, which is a severe limitation for quantum machine learning algorithms. Quantum Splines (QSplines) have recently been proposed to approximate quantum activation functions to introduce non-linearity in quantum algorithms. However, QSplines make use of the HHL as a subroutine and require a fault-tolerant quantum computer to be correctly implemented. This work proposes the Generalised QSplines (GQSplines), a novel method for approximating non-linear quantum activation functions using hybrid quantum-classical computation. The GQSplines overcome the highly demanding requirements of the original QSplines in terms of quantum hardware and can be implemented using near-term quantum computers. Furthermore, the proposed method relies on a flexible problem representation for non-linear approximation and it is suitable to be embedded in existing quantum neural network architectures. In addition, we provide a practical implementation of GQSplines using Pennylane and show that our model outperforms the original QSplines in terms of quality of fitting.
☆ Fourier-MIONet: Fourier-enhanced multiple-input neural operators for multiphase modeling of geological carbon sequestration
Geologic Carbon Storage (GCS) is an important technology that aims to reduce the amount of carbon dioxide in the atmosphere. Multiphase flow in porous media is essential to understand CO2 migration and pressure fields in the subsurface associated with GCS. However, numerical simulation for such problems in 4D is computationally challenging and expensive, due to the multiphysics and multiscale nature of the highly nonlinear governing partial differential equations (PDEs). It prevents us from considering multiple subsurface scenarios and conducting real-time optimization. Here, we develop a Fourier-enhanced multiple-input neural operator (Fourier-MIONet) to learn the solution operator of the problem of multiphase flow in porous media. Fourier-MIONet utilizes the recently developed framework of the multiple-input deep neural operators (MIONet) and incorporates the Fourier neural operator (FNO) in the network architecture. Once Fourier-MIONet is trained, it can predict the evolution of saturation and pressure of the multiphase flow under various reservoir conditions, such as permeability and porosity heterogeneity, anisotropy, injection configurations, and multiphase flow properties. Compared to the enhanced FNO (U-FNO), the proposed Fourier-MIONet has 90% fewer unknown parameters, and it can be trained in significantly less time (about 3.5 times faster) with much lower CPU memory (< 15%) and GPU memory (< 35%) requirements, to achieve similar prediction accuracy. In addition to the lower computational cost, Fourier-MIONet can be trained with only 6 snapshots of time to predict the PDE solutions for 30 years. The excellent generalizability of Fourier-MIONet is enabled by its adherence to the physical principle that the solution to a PDE is continuous over time.
☆ LMI-based Data-Driven Robust Model Predictive Control
Predictive control, which is based on a model of the system to compute the applied input optimizing the future system behavior, is by now widely used. If the nominal models are not given or are very uncertain, data-driven model predictive control approaches can be employed, where the system model or input is directly obtained from past measured trajectories. Using a data informativity framework and Finsler's lemma, we propose a data-driven robust linear matrix inequality-based model predictive control scheme that considers input and state constraints. Using these data, we formulate the problem as a semi-definite optimization problem, whose solution provides the matrix gain for the linear feedback, while the decisive variables are independent of the length of the measurement data. The designed controller stabilizes the closed-loop system asymptotically and guarantees constraint satisfaction. Numerical examples are conducted to illustrate the method.
☆ Multilevel Diffusion: Infinite Dimensional Score-Based Diffusion Models for Image Generation
Score-based diffusion models (SBDM) have recently emerged as state-of-the-art approaches for image generation. Existing SBDMs are typically formulated in a finite-dimensional setting, where images are considered as tensors of a finite size. This papers develops SBDMs in the infinite-dimensional setting, that is, we model the training data as functions supported on a rectangular domain. Besides the quest for generating images at ever higher resolution our primary motivation is to create a well-posed infinite-dimensional learning problem so that we can discretize it consistently on multiple resolution levels. We thereby hope to obtain diffusion models that generalize across different resolution levels and improve the efficiency of the training process. We demonstrate how to overcome two shortcomings of current SBDM approaches in the infinite-dimensional setting. First, we modify the forward process to ensure that the latent distribution is well-defined in the infinite-dimensional setting using the notion of trace class operators. Second, we illustrate that approximating the score function with an operator network, in our case Fourier neural operators (FNOs), is beneficial for multilevel training. After deriving the forward and reverse process in the infinite-dimensional setting, we show their well-posedness, derive adequate discretizations, and investigate the role of the latent distributions. We provide first promising numerical results on two datasets, MNIST and material structures. In particular, we show that multilevel training is feasible within this framework.
☆ FastFill: Efficient Compatible Model Update
In many retrieval systems the original high dimensional data (e.g., images) is mapped to a lower dimensional feature through a learned embedding model. The task of retrieving the most similar data from a gallery set to a given query data is performed through a similarity comparison on features. When the embedding model is updated, it might produce features that are not comparable/compatible with features already in the gallery computed with the old model. Subsequently, all features in the gallery need to be re-computed using the new embedding model -- a computationally expensive process called backfilling. Recently, compatible representation learning methods have been proposed to avoid backfilling. Despite their relative success, there is an inherent trade-off between the new model performance and its compatibility with the old model. In this work, we introduce FastFill: a compatible model update process using feature alignment and policy based partial backfilling to promptly elevate retrieval performance. We show that previous backfilling strategies suffer from decreased performance and demonstrate the importance of both the training objective and the ordering in online partial backfilling. We propose a new training method for feature alignment between old and new embedding models using uncertainty estimation. Compared to previous works, we obtain significantly improved backfilling results on a variety of datasets: mAP on ImageNet (+4.4\%), Places-365 (+2.7\%), and VGG-Face2 (+1.3\%). Further, we demonstrate that when updating a biased model with FastFill, the minority subgroup accuracy gap promptly vanishes with a small fraction of partial backfilling.
comment: To appear in The Eleventh International Conference on Learning Representations
☆ RAF: Holistic Compilation for Deep Learning Model Training
As deep learning is pervasive in modern applications, many deep learning frameworks are presented for deep learning practitioners to develop and train DNN models rapidly. Meanwhile, as training large deep learning models becomes a trend in recent years, the training throughput and memory footprint are getting crucial. Accordingly, optimizing training workloads with compiler optimizations is inevitable and getting more and more attentions. However, existing deep learning compilers (DLCs) mainly target inference and do not incorporate holistic optimizations, such as automatic differentiation and automatic mixed precision, in training workloads. In this paper, we present RAF, a deep learning compiler for training. Unlike existing DLCs, RAF accepts a forward model and in-house generates a training graph. Accordingly, RAF is able to systematically consolidate graph optimizations for performance, memory and distributed training. In addition, to catch up to the state-of-the-art performance with hand-crafted kernel libraries as well as tensor compilers, RAF proposes an operator dialect mechanism to seamlessly integrate all possible kernel implementations. We demonstrate that by in-house training graph generation and operator dialect mechanism, we are able to perform holistic optimizations and achieve either better training throughput or larger batch size against PyTorch (eager and torchscript mode), XLA, and DeepSpeed for popular transformer models on GPUs.
☆ Meta-learning Control Variates: Variance Reduction with Limited Data
Control variates can be a powerful tool to reduce the variance of Monte Carlo estimators, but constructing effective control variates can be challenging when the number of samples is small. In this paper, we show that when a large number of related integrals need to be computed, it is possible to leverage the similarity between these integration tasks to improve performance even when the number of samples per task is very small. Our approach, called meta learning CVs (Meta-CVs), can be used for up to hundreds or thousands of tasks. Our empirical assessment indicates that Meta-CVs can lead to significant variance reduction in such settings, and our theoretical analysis establishes general conditions under which Meta-CVs can be successfully trained.
☆ A General Theory of Correct, Incorrect, and Extrinsic Equivariance
Although equivariant machine learning has proven effective at many tasks, success depends heavily on the assumption that the ground truth function is symmetric over the entire domain matching the symmetry in an equivariant neural network. A missing piece in the equivariant learning literature is the analysis of equivariant networks when symmetry exists only partially in the domain. In this work, we present a general theory for such a situation. We propose pointwise definitions of correct, incorrect, and extrinsic equivariance, which allow us to quantify continuously the degree of each type of equivariance a function displays. We then study the impact of various degrees of incorrect or extrinsic symmetry on model error. We prove error lower bounds for invariant or equivariant networks in classification or regression settings with partially incorrect symmetry. We also analyze the potentially harmful effects of extrinsic equivariance. Experiments validate these results in three different environments.
☆ Vector Quantized Time Series Generation with a Bidirectional Prior Model AISTATS 2023
Time series generation (TSG) studies have mainly focused on the use of Generative Adversarial Networks (GANs) combined with recurrent neural network (RNN) variants. However, the fundamental limitations and challenges of training GANs still remain. In addition, the RNN-family typically has difficulties with temporal consistency between distant timesteps. Motivated by the successes in the image generation (IMG) domain, we propose TimeVQVAE, the first work, to our knowledge, that uses vector quantization (VQ) techniques to address the TSG problem. Moreover, the priors of the discrete latent spaces are learned with bidirectional transformer models that can better capture global temporal consistency. We also propose VQ modeling in a time-frequency domain, separated into low-frequency (LF) and high-frequency (HF). This allows us to retain important characteristics of the time series and, in turn, generate new synthetic signals that are of better quality, with sharper changes in modularity, than its competing TSG methods. Our experimental evaluation is conducted on all datasets from the UCR archive, using well-established metrics in the IMG literature, such as Fr\'echet inception distance and inception scores. Our implementation on GitHub: \url{https://github.com/ML4ITS/TimeVQVAE}.
comment: accepted at AISTATS 2023
☆ Advancing Direct Convolution using Convolution Slicing Optimization and ISA Extensions
Convolution is one of the most computationally intensive operations that must be performed for machine-learning model inference. A traditional approach to compute convolutions is known as the Im2Col + BLAS method. This paper proposes SConv: a direct-convolution algorithm based on a MLIR/LLVM code-generation toolchain that can be integrated into machine-learning compilers . This algorithm introduces: (a) Convolution Slicing Analysis (CSA) - a convolution-specific 3D cache-blocking analysis pass that focuses on tile reuse over the cache hierarchy; (b) Convolution Slicing Optimization (CSO) - a code-generation pass that uses CSA to generate a tiled direct-convolution macro-kernel; and (c) Vector-Based Packing (VBP) - an architecture-specific optimized input-tensor packing solution based on vector-register shift instructions for convolutions with unitary stride. Experiments conducted on 393 convolutions from full ONNX-MLIR machine-learning models indicate that the elimination of the Im2Col transformation and the use of fast packing routines result in a total packing time reduction, on full model inference, of 2.0x - 3.9x on Intel x86 and 3.6x - 7.2x on IBM POWER10. The speed-up over an Im2Col + BLAS method based on current BLAS implementations for end-to-end machine-learning model inference is in the range of 9% - 25% for Intel x86 and 10% - 42% for IBM POWER10 architectures. The total convolution speedup for model inference is 12% - 27% on Intel x86 and 26% - 46% on IBM POWER10. SConv also outperforms BLAS GEMM, when computing pointwise convolutions, in more than 83% of the 219 tested instances.
comment: 15 pages, 11 figures
☆ On the Risks of Stealing the Decoding Algorithms of Language Models
A key component of generating text from modern language models (LM) is the selection and tuning of decoding algorithms. These algorithms determine how to generate text from the internal probability distribution generated by the LM. The process of choosing a decoding algorithm and tuning its hyperparameters takes significant time, manual effort, and computation, and it also requires extensive human evaluation. Therefore, the identity and hyperparameters of such decoding algorithms are considered to be extremely valuable to their owners. In this work, we show, for the first time, that an adversary with typical API access to an LM can steal the type and hyperparameters of its decoding algorithms at very low monetary costs. Our attack is effective against popular LMs used in text generation APIs, including GPT-2 and GPT-3. We demonstrate the feasibility of stealing such information with only a few dollars, e.g., $\$0.8$, $\$1$, $\$4$, and $\$40$ for the four versions of GPT-3.
☆ Model Predictive Control with Gaussian-Process-Supported Dynamical Constraints for Autonomous Vehicles
We propose a model predictive control approach for autonomous vehicles that exploits learned Gaussian processes for predicting human driving behavior. The proposed approach employs the uncertainty about the GP's prediction to achieve safety. A multi-mode predictive control approach considers the possible intentions of the human drivers. While the intentions are represented by different Gaussian processes, their probabilities foreseen in the observed behaviors are determined by a suitable online classification. Intentions below a certain probability threshold are neglected to improve performance. The proposed multi-mode model predictive control approach with Gaussian process regression support enables repeated feasibility and probabilistic constraint satisfaction with high probability. The approach is underlined in simulation, considering real-world measurements for training the Gaussian processes.
☆ Fast offset corrected in-memory training
In-memory computing with resistive crossbar arrays has been suggested to accelerate deep-learning workloads in highly efficient manner. To unleash the full potential of in-memory computing, it is desirable to accelerate the training as well as inference for large deep neural networks (DNNs). In the past, specialized in-memory training algorithms have been proposed that not only accelerate the forward and backward passes, but also establish tricks to update the weight in-memory and in parallel. However, the state-of-the-art algorithm (Tiki-Taka version 2 (TTv2)) still requires near perfect offset correction and suffers from potential biases that might occur due to programming and estimation inaccuracies, as well as longer-term instabilities of the device materials. Here we propose and describe two new and improved algorithms for in-memory computing (Chopped-TTv2 (c-TTv2) and Analog Gradient Accumulation with Dynamic reference (AGAD)), that retain the same runtime complexity but correct for any remaining offsets using choppers. These algorithms greatly relax the device requirements and thus expanding the scope of possible materials potentially employed for such fast in-memory DNN training.
comment: 14 pages, 10 figures
☆ VOLTA: an Environment-Aware Contrastive Cell Representation Learning for Histopathology
In clinical practice, many diagnosis tasks rely on the identification of cells in histopathology images. While supervised machine learning techniques require labels, providing manual cell annotations is time-consuming due to the large number of cells. In this paper, we propose a self-supervised framework (VOLTA) for cell representation learning in histopathology images using a novel technique that accounts for the cell's mutual relationship with its environment for improved cell representations. We subjected our model to extensive experiments on the data collected from multiple institutions around the world comprising of over 700,000 cells, four cancer types, and cell types ranging from three to six categories for each dataset. The results show that our model outperforms the state-of-the-art models in cell representation learning. To showcase the potential power of our proposed framework, we applied VOLTA to ovarian and endometrial cancers with very small sample sizes (10-20 samples) and demonstrated that our cell representations can be utilized to identify the known histotypes of ovarian cancer and provide novel insights that link histopathology and molecular subtypes of endometrial cancer. Unlike supervised deep learning models that require large sample sizes for training, we provide a framework that can empower new discoveries without any annotation data in situations where sample sizes are limited.
☆ A path in regression Random Forest looking for spatial dependence: a taxonomy and a systematic review
Random Forest (RF) is a well-known data-driven algorithm applied in several fields thanks to its flexibility in modeling the relationship between the response variable and the predictors, also in case of strong non-linearities. In environmental applications, it often occurs that the phenomenon of interest may present spatial and/or temporal dependence that is not taken explicitly into account by RF in its standard version. In this work, we propose a taxonomy to classify strategies according to when (Pre-, In- and/or Post-processing) they try to include the spatial information into regression RF. Moreover, we provide a systematic review and classify the most recent strategies adopted to "adjust" regression RF to spatially dependent data, based on the criteria provided by the Preferred Reporting Items for Systematic reviews and Meta-Analysis (PRISMA). The latter consists of a reproducible methodology for collecting and processing existing literature on a specified topic from different sources. PRISMA starts with a query and ends with a set of scientific documents to review: we performed an online query on the 25$^{th}$ October 2022 and, in the end, 32 documents were considered for review. The employed methodological strategies and the application fields considered in the 32 scientific documents are described and discussed.
☆ Considerations on the Theory of Training Models with Differential Privacy
In federated learning collaborative learning takes place by a set of clients who each want to remain in control of how their local training data is used, in particular, how can each client's local training data remain private? Differential privacy is one method to limit privacy leakage. We provide a general overview of its framework and provable properties, adopt the more recent hypothesis based definition called Gaussian DP or $f$-DP, and discuss Differentially Private Stochastic Gradient Descent (DP-SGD). We stay at a meta level and attempt intuitive explanations and insights \textit{in this book chapter}.
comment: 18 pages, a book chapter. arXiv admin note: text overlap with arXiv:2212.05796
☆ Cost-Effective Hyperparameter Optimization for Large Language Model Generation Inference
Large Language Models (LLMs) like GPT-3 have sparked significant interest in their generative capabilities, leading to the development of various commercial applications. The high cost of using the models drives application builders to maximize the value of generation under a limited inference budget. This paper presents a study of optimizing inference hyperparameters like the number of responses, temperature and max tokens, which significantly affects the utility/cost of text generation. We design a framework named EcoOptiGen which leverages economical hyperparameter optimization and cost-based pruning. Experiments with the latest GPT-3.5 models on a variety of tasks verify its effectiveness. EcoOptiGen is implemented in the FLAML library: https://github.com/microsoft/FLAML, and we provide one example of using it at: https://microsoft.github.io/FLAML/docs/Examples/Integrate%20-%20OpenAI.
☆ Neural Probabilistic Logic Programming in Discrete-Continuous Domains
Neural-symbolic AI (NeSy) allows neural networks to exploit symbolic background knowledge in the form of logic. It has been shown to aid learning in the limited data regime and to facilitate inference on out-of-distribution data. Probabilistic NeSy focuses on integrating neural networks with both logic and probability theory, which additionally allows learning under uncertainty. A major limitation of current probabilistic NeSy systems, such as DeepProbLog, is their restriction to finite probability distributions, i.e., discrete random variables. In contrast, deep probabilistic programming (DPP) excels in modelling and optimising continuous probability distributions. Hence, we introduce DeepSeaProbLog, a neural probabilistic logic programming language that incorporates DPP techniques into NeSy. Doing so results in the support of inference and learning of both discrete and continuous probability distributions under logical constraints. Our main contributions are 1) the semantics of DeepSeaProbLog and its corresponding inference algorithm, 2) a proven asymptotically unbiased learning algorithm, and 3) a series of experiments that illustrate the versatility of our approach.
comment: 27 pages, 9 figures
☆ MCTS-GEB: Monte Carlo Tree Search is a Good E-graph Builder
Rewrite systems [6, 10, 12] have been widely employing equality saturation [9], which is an optimisation methodology that uses a saturated e-graph to represent all possible sequences of rewrite simultaneously, and then extracts the optimal one. As such, optimal results can be achieved by avoiding the phase-ordering problem. However, we observe that when the e-graph is not saturated, it cannot represent all possible rewrite opportunities and therefore the phase-ordering problem is re-introduced during the construction phase of the e-graph. To address this problem, we propose MCTS-GEB, a domain-general rewrite system that applies reinforcement learning (RL) to e-graph construction. At its core, MCTS-GEB uses a Monte Carlo Tree Search (MCTS) [3] to efficiently plan for the optimal e-graph construction, and therefore it can effectively eliminate the phase-ordering problem at the construction phase and achieve better performance within a reasonable time. Evaluation in two different domains shows MCTS-GEB can outperform the state-of-the-art rewrite systems by up to 49x, while the optimisation can generally take less than an hour, indicating MCTS-GEB is a promising building block for the future generation of rewrite systems.
☆ Forecasting the movements of Bitcoin prices: an application of machine learning algorithms
Cryptocurrencies, such as Bitcoin, are one of the most controversial and complex technological innovations in today's financial system. This study aims to forecast the movements of Bitcoin prices at a high degree of accuracy. To this aim, four different Machine Learning (ML) algorithms are applied, namely, the Support Vector Machines (SVM), the Artificial Neural Network (ANN), the Naive Bayes (NB) and the Random Forest (RF) besides the logistic regression (LR) as a benchmark model. In order to test these algorithms, besides existing continuous dataset, discrete dataset was also created and used. For the evaluations of algorithm performances, the F statistic, accuracy statistic, the Mean Absolute Error (MAE), the Root Mean Square Error (RMSE) and the Root Absolute Error (RAE) metrics were used. The t test was used to compare the performances of the SVM, ANN, NB and RF with the performance of the LR. Empirical findings reveal that, while the RF has the highest forecasting performance in the continuous dataset, the NB has the lowest. On the other hand, while the ANN has the highest and the NB the lowest performance in the discrete dataset. Furthermore, the discrete dataset improves the overall forecasting performance in all algorithms (models) estimated.
comment: 14 pages, 2 figures and 15 tables
☆ Robust Multimodal Fusion for Human Activity Recognition
The proliferation of IoT and mobile devices equipped with heterogeneous sensors has enabled new applications that rely on the fusion of time-series data generated by multiple sensors with different modalities. While there are promising deep neural network architectures for multimodal fusion, their performance falls apart quickly in the presence of consecutive missing data and noise across multiple modalities/sensors, the issues that are prevalent in real-world settings. We propose Centaur, a multimodal fusion model for human activity recognition (HAR) that is robust to these data quality issues. Centaur combines a data cleaning module, which is a denoising autoencoder with convolutional layers, and a multimodal fusion module, which is a deep convolutional neural network with the self-attention mechanism to capture cross-sensor correlation. We train Centaur using a stochastic data corruption scheme and evaluate it on three datasets that contain data generated by multiple inertial measurement units. Centaur's data cleaning module outperforms 2 state-of-the-art autoencoder-based models and its multimodal fusion module outperforms 4 strong baselines. Compared to 2 related robust fusion architectures, Centaur is more robust, achieving 11.59-17.52% higher accuracy in HAR, especially in the presence of consecutive missing data in multiple sensor channels.
comment: 13 pages
☆ Diffusing Gaussian Mixtures for Generating Categorical Data
Learning a categorical distribution comes with its own set of challenges. A successful approach taken by state-of-the-art works is to cast the problem in a continuous domain to take advantage of the impressive performance of the generative models for continuous data. Amongst them are the recently emerging diffusion probabilistic models, which have the observed advantage of generating high-quality samples. Recent advances for categorical generative models have focused on log likelihood improvements. In this work, we propose a generative model for categorical data based on diffusion models with a focus on high-quality sample generation, and propose sampled-based evaluation methods. The efficacy of our method stems from performing diffusion in the continuous domain while having its parameterization informed by the structure of the categorical nature of the target distribution. Our method of evaluation highlights the capabilities and limitations of different generative models for generating categorical data, and includes experiments on synthetic and real-world protein datasets.
☆ Contribution of clinical course to outcome after traumatic brain injury: mining patient trajectories from European intensive care unit data
Existing methods to characterise the evolving condition of traumatic brain injury (TBI) patients in the intensive care unit (ICU) do not capture the context necessary for individualising treatment. We aimed to develop a modelling strategy which integrates all data stored in medical records to produce an interpretable disease course for each TBI patient's ICU stay. From a prospective, European cohort (n=1,550, 65 centres, 19 countries) of TBI patients, we extracted all 1,166 variables collected before or during ICU stay as well as 6-month functional outcome on the Glasgow Outcome Scale-Extended (GOSE). We trained recurrent neural network models to map a token-embedded time series representation of all variables (including missing data) to an ordinal GOSE prognosis every 2 hours. With repeated cross-validation, we evaluated calibration and the explanation of ordinal variance in GOSE with Somers' Dxy. Furthermore, we applied TimeSHAP to calculate the contribution of variables and prior timepoints towards transitions in patient trajectories. Our modelling strategy achieved calibration at 8 hours, and the full range of variables explained up to 52% (95% CI: 50-54%) of the variance in ordinal functional outcome. Up to 91% (90-91%) of this explanation was derived from pre-ICU and admission information. Information collected in the ICU increased explanation (by up to 5% [4-6%]), though not enough to counter poorer performance in longer-stay (>5.75 days) patients. Static variables with the highest contributions were physician prognoses and certain demographic and CT features. Among dynamic variables, markers of intracranial hypertension and neurological function contributed the most. Whilst static information currently accounts for the majority of functional outcome explanation, our data-driven analysis highlights investigative avenues to improve dynamic characterisation of longer-stay patients.
☆ ELF: Federated Langevin Algorithms with Primal, Dual and Bidirectional Compression
Federated sampling algorithms have recently gained great popularity in the community of machine learning and statistics. This paper studies variants of such algorithms called Error Feedback Langevin algorithms (ELF). In particular, we analyze the combinations of EF21 and EF21-P with the federated Langevin Monte-Carlo. We propose three algorithms: P-ELF, D-ELF, and B-ELF that use, respectively, primal, dual, and bidirectional compressors. We analyze the proposed methods under Log-Sobolev inequality and provide non-asymptotic convergence guarantees.
☆ Densely Connected $G$-invariant Deep Neural Networks with Signed Permutation Representations
We introduce and investigate, for finite groups $G$, $G$-invariant deep neural network ($G$-DNN) architectures with ReLU activation that are densely connected -- i.e., include all possible skip connections. In contrast to other $G$-invariant architectures in the literature, the preactivations of the$G$-DNNs presented here are able to transform by \emph{signed} permutation representations (signed perm-reps) of $G$. Moreover, the individual layers of the $G$-DNNs are not required to be $G$-equivariant; instead, the preactivations are constrained to be $G$-equivariant functions of the network input in a way that couples weights across all layers. The result is a richer family of $G$-invariant architectures never seen previously. We derive an efficient implementation of $G$-DNNs after a reparameterization of weights, as well as necessary and sufficient conditions for an architecture to be "admissible" -- i.e., nondegenerate and inequivalent to smaller architectures. We include code that allows a user to build a $G$-DNN interactively layer-by-layer, with the final architecture guaranteed to be admissible. Finally, we apply $G$-DNNs to two example problems -- (1) multiplication in $\{-1, 1\}$ (with theoretical guarantees) and (2) 3D object classification -- finding that the inclusion of signed perm-reps significantly boosts predictive performance compared to baselines with only ordinary (i.e., unsigned) perm-reps.
comment: 33 pages, 2 figures. For associated code repository see https://github.com/dagrawa2/gdnn_code
☆ The Descriptive Complexity of Graph Neural Networks
We analyse the power of graph neural networks (GNNs) in terms of Boolean circuit complexity and descriptive complexity. We prove that the graph queries that can be computed by a polynomial-size bounded-depth family of GNNs are exactly those definable in the guarded fragment GFO+C of first-order logic with counting and with built-in relations. This puts GNNs in the circuit complexity class TC^0. Remarkably, the GNN families may use arbitrary real weights and a wide class of activation functions that includes the standard ReLU, logistic "sigmoid", and hyperbolic tangent functions. If the GNNs are allowed to use random initialisation and global readout (both standard features of GNNs widely used in practice), they can compute exactly the same queries as bounded depth Boolean circuits with threshold gates, that is, exactly the queries in TC^0. Moreover, we show that queries computable by a single GNN with piecewise linear activations and rational weights are definable in GFO+C without built-in relations. Therefore, they are contained in uniform TC^0.
☆ Differential Privacy Meets Neural Network Pruning
A major challenge in applying differential privacy to training deep neural network models is scalability.The widely-used training algorithm, differentially private stochastic gradient descent (DP-SGD), struggles with training moderately-sized neural network models for a value of epsilon corresponding to a high level of privacy protection. In this paper, we explore the idea of dimensionality reduction inspired by neural network pruning to improve the scalability of DP-SGD. We study the interplay between neural network pruning and differential privacy, through the two modes of parameter updates. We call the first mode, parameter freezing, where we pre-prune the network and only update the remaining parameters using DP-SGD. We call the second mode, parameter selection, where we select which parameters to update at each step of training and update only those selected using DP-SGD. In these modes, we use public data for freezing or selecting parameters to avoid privacy loss incurring in these steps. Naturally, the closeness between the private and public data plays an important role in the success of this paradigm. Our experimental results demonstrate how decreasing the parameter space improves differentially private training. Moreover, by studying two popular forms of pruning which do not rely on gradients and do not incur an additional privacy loss, we show that random selection performs on par with magnitude-based selection when it comes to DP-SGD training.
☆ Application of supervised learning models in the Chinese futures market
Based on the characteristics of the Chinese futures market, this paper builds a supervised learning model to predict the trend of futures prices and then designs a trading strategy based on the prediction results. The Precision, Recall and F1-score of the classification problem show that our model can meet the accuracy requirements for the classification of futures price movements in terms of test data. The backtest results show that our trading system has an upward trending return curve with low capital retracement.
☆ "How to make them stay?" -- Diverse Counterfactual Explanations of Employee Attrition
Employee attrition is an important and complex problem that can directly affect an organisation's competitiveness and performance. Explaining the reasons why employees leave an organisation is a key human resource management challenge due to the high costs and time required to attract and keep talented employees. Businesses therefore aim to increase employee retention rates to minimise their costs and maximise their performance. Machine learning (ML) has been applied in various aspects of human resource management including attrition prediction to provide businesses with insights on proactive measures on how to prevent talented employees from quitting. Among these ML methods, the best performance has been reported by ensemble or deep neural networks, which by nature constitute black box techniques and thus cannot be easily interpreted. To enable the understanding of these models' reasoning several explainability frameworks have been proposed. Counterfactual explanation methods have attracted considerable attention in recent years since they can be used to explain and recommend actions to be performed to obtain the desired outcome. However current counterfactual explanations methods focus on optimising the changes to be made on individual cases to achieve the desired outcome. In the attrition problem it is important to be able to foresee what would be the effect of an organisation's action to a group of employees where the goal is to prevent them from leaving the company. Therefore, in this paper we propose the use of counterfactual explanations focusing on multiple attrition cases from historical data, to identify the optimum interventions that an organisation needs to make to its practices/policies to prevent or minimise attrition probability for these cases.
comment: Accepted as a short paper at ICEIS 2023
☆ Safe Machine-Learning-supported Model Predictive Force and Motion Control in Robotics
Many robotic tasks, such as human-robot interactions or the handling of fragile objects, require tight control and limitation of appearing forces and moments alongside sensible motion control to achieve safe yet high-performance operation. We propose a learning-supported model predictive force and motion control scheme that provides stochastic safety guarantees while adapting to changing situations. Gaussian processes are used to learn the uncertain relations that map the robot's states to the forces and moments. The model predictive controller uses these Gaussian process models to achieve precise motion and force control under stochastic constraint satisfaction. As the uncertainty only occurs in the static model parts -- the output equations -- a computationally efficient stochastic MPC formulation is used. Analysis of recursive feasibility of the optimal control problem and convergence of the closed loop system for the static uncertainty case are given. Chance constraint formulation and back-offs are constructed based on the variance of the Gaussian process to guarantee safe operation. The approach is illustrated on a lightweight robot in simulations and experiments.
☆ Extrapolative Controlled Sequence Generation via Iterative Refinement
We study the problem of extrapolative controlled generation, i.e., generating sequences with attribute values beyond the range seen in training. This task is of significant importance in automated design, especially drug discovery, where the goal is to design novel proteins that are \textit{better} (e.g., more stable) than existing sequences. Thus, by definition, the target sequences and their attribute values are out of the training distribution, posing challenges to existing methods that aim to directly generate the target sequence. Instead, in this work, we propose Iterative Controlled Extrapolation (ICE) which iteratively makes local edits to a sequence to enable extrapolation. We train the model on synthetically generated sequence pairs that demonstrate small improvement in the attribute value. Results on one natural language task (sentiment analysis) and two protein engineering tasks (ACE2 stability and AAV fitness) show that ICE considerably outperforms state-of-the-art approaches despite its simplicity. Our code and models are available at: https://github.com/vishakhpk/iter-extrapolation.
comment: Preprint
☆ Byzantine-Robust Loopless Stochastic Variance-Reduced Gradient
Distributed optimization with open collaboration is a popular field since it provides an opportunity for small groups/companies/universities, and individuals to jointly solve huge-scale problems. However, standard optimization algorithms are fragile in such settings due to the possible presence of so-called Byzantine workers -- participants that can send (intentionally or not) incorrect information instead of the one prescribed by the protocol (e.g., send anti-gradient instead of stochastic gradients). Thus, the problem of designing distributed methods with provable robustness to Byzantine workers has been receiving a lot of attention recently. In particular, several works consider a very promising way to achieve Byzantine tolerance via exploiting variance reduction and robust aggregation. The existing approaches use SAGA- and SARAH-type variance-reduced estimators, while another popular estimator -- SVRG -- is not studied in the context of Byzantine-robustness. In this work, we close this gap in the literature and propose a new method -- Byzantine-Robust Loopless Stochastic Variance Reduced Gradient (BR-LSVRG). We derive non-asymptotic convergence guarantees for the new method in the strongly convex case and compare its performance with existing approaches in numerical experiments.
comment: 15 pages, 2 figures. Code: https://github.com/Nikosimus/BR-LSVRG
☆ Streaming Kernel PCA Algorithm With Small Space
Principal Component Analysis (PCA) is a widely used technique in machine learning, data analysis and signal processing. With the increase in the size and complexity of datasets, it has become important to develop low-space usage algorithms for PCA. Streaming PCA has gained significant attention in recent years, as it can handle large datasets efficiently. The kernel method, which is commonly used in learning algorithms such as Support Vector Machines (SVMs), has also been applied in PCA algorithms. We propose a streaming algorithm for Kernel PCA problems based on the traditional scheme by Oja. Our algorithm addresses the challenge of reducing the memory usage of PCA while maintaining its accuracy. We analyze the performance of our algorithm by studying the conditions under which it succeeds. Specifically, we show that, when the spectral ratio $R := \lambda_1/\lambda_2$ of the target covariance matrix is lower bounded by $C \cdot \log n\cdot \log d$, the streaming PCA can be solved with $O(d)$ space cost. Our proposed algorithm has several advantages over existing methods. First, it is a streaming algorithm that can handle large datasets efficiently. Second, it employs the kernel method, which allows it to capture complex nonlinear relationships among data points. Third, it has a low-space usage, making it suitable for applications where memory is limited.
☆ Sketching with Spherical Designs for Noisy Data Fitting on Spheres
This paper proposes a sketching strategy based on spherical designs, which is applied to the classical spherical basis function approach for massive spherical data fitting. We conduct theoretical analysis and numerical verifications to demonstrate the feasibility of the proposed { sketching} strategy. From the theoretical side, we prove that sketching based on spherical designs can reduce the computational burden of the spherical basis function approach without sacrificing its approximation capability. In particular, we provide upper and lower bounds for the proposed { sketching} strategy to fit noisy data on spheres. From the experimental side, we numerically illustrate the feasibility of the sketching strategy by showing its comparable fitting performance with the spherical basis function approach. These interesting findings show that the proposed sketching strategy is capable of fitting massive and noisy data on spheres.
☆ Unimodal Distributions for Ordinal Regression
In many real-world prediction tasks, class labels contain information about the relative order between labels that are not captured by commonly used loss functions such as multicategory cross-entropy. Recently, the preference for unimodal distributions in the output space has been incorporated into models and loss functions to account for such ordering information. However, current approaches rely on heuristics that lack a theoretical foundation. Here, we propose two new approaches to incorporate the preference for unimodal distributions into the predictive model. We analyse the set of unimodal distributions in the probability simplex and establish fundamental properties. We then propose a new architecture that imposes unimodal distributions and a new loss term that relies on the notion of projection in a set to promote unimodality. Experiments show the new architecture achieves top-2 performance, while the proposed new loss term is very competitive while maintaining high unimodality.
comment: 6 pages
☆ A robust method for reliability updating with equality information using sequential adaptive importance sampling
Reliability updating refers to a problem that integrates Bayesian updating technique with structural reliability analysis and cannot be directly solved by structural reliability methods (SRMs) when it involves equality information. The state-of-the-art approaches transform equality information into inequality information by introducing an auxiliary standard normal parameter. These methods, however, encounter the loss of computational efficiency due to the difficulty in finding the maximum of the likelihood function, the large coefficient of variation (COV) associated with the posterior failure probability and the inapplicability to dynamic updating problems where new information is constantly available. To overcome these limitations, this paper proposes an innovative method called RU-SAIS (reliability updating using sequential adaptive importance sampling), which combines elements of sequential importance sampling and K-means clustering to construct a series of important sampling densities (ISDs) using Gaussian mixture. The last ISD of the sequence is further adaptively modified through application of the cross entropy method. The performance of RU-SAIS is demonstrated by three examples. Results show that RU-SAIS achieves a more accurate and robust estimator of the posterior failure probability than the existing methods such as subset simulation.
comment: 38 pages, 6 tables, 9 figures
☆ Magnushammer: A Transformer-based Approach to Premise Selection
Premise selection is a fundamental problem of automated theorem proving. Previous works often use intricate symbolic methods, rely on domain knowledge, and require significant engineering effort to solve this task. In this work, we show that Magnushammer, a neural transformer-based approach, can outperform traditional symbolic systems by a large margin. Tested on the PISA benchmark, Magnushammer achieves $59.5\%$ proof rate compared to a $38.3\%$ proof rate of Sledgehammer, the most mature and popular symbolic-based solver. Furthermore, by combining Magnushammer with a neural formal prover based on a language model, we significantly improve the previous state-of-the-art proof rate from $57.0\%$ to $71.0\%$.
☆ Better Together: Using Multi-task Learning to Improve Feature Selection within Structural Datasets
There have been recent efforts to move to population-based structural health monitoring (PBSHM) systems. One area of PBSHM which has been recognised for potential development is the use of multi-task learning (MTL); algorithms which differ from traditional independent learning algorithms. Presented here is the use of the MTL, ''Joint Feature Selection with LASSO'', to provide automatic feature selection for a structural dataset. The classification task is to differentiate between the port and starboard side of a tailplane, for samples from two aircraft of the same model. The independent learner produced perfect F1 scores but had poor engineering insight; whereas the MTL results were interpretable, highlighting structural differences as opposed to differences in experimental set-up.
☆ Extracting Digital Biomarkers for Unobtrusive Stress State Screening from Multimodal Wearable Data
With the development of wearable technologies, a new kind of healthcare data has become valuable as medical information. These data provide meaningful information regarding an individual's physiological and psychological states, such as activity level, mood, stress, and cognitive health. These biomarkers are named digital since they are collected from digital devices integrated with various sensors. In this study, we explore digital biomarkers related to stress modality by examining data collected from mobile phones and smartwatches. We utilize machine learning techniques on the Tesserae dataset, precisely Random Forest, to extract stress biomarkers. Using feature selection techniques, we utilize weather, activity, heart rate (HR), stress, sleep, and location (work-home) measurements from wearables to determine the most important stress-related biomarkers. We believe we contribute to interpreting stress biomarkers with a high range of features from different devices. In addition, we classify the $5$ different stress levels with the most important features, and our results show that we can achieve $85\%$ overall class accuracy by adjusting class imbalance and adding extra features related to personality characteristics. We perform similar and even better results in recognizing stress states with digital biomarkers in a daily-life scenario targeting a higher number of classes compared to the related studies.
☆ Graph Neural Networks Enhanced Smart Contract Vulnerability Detection of Educational Blockchain
With the development of blockchain technology, more and more attention has been paid to the intersection of blockchain and education, and various educational evaluation systems and E-learning systems are developed based on blockchain technology. Among them, Ethereum smart contract is favored by developers for its ``event-triggered" mechanism for building education intelligent trading systems and intelligent learning platforms. However, due to the immutability of blockchain, published smart contracts cannot be modified, so problematic contracts cannot be fixed by modifying the code in the educational blockchain. In recent years, security incidents due to smart contract vulnerabilities have caused huge property losses, so the detection of smart contract vulnerabilities in educational blockchain has become a great challenge. To solve this problem, this paper proposes a graph neural network (GNN) based vulnerability detection for smart contracts in educational blockchains. Firstly, the bytecodes are decompiled to get the opcode. Secondly, the basic blocks are divided, and the edges between the basic blocks according to the opcode execution logic are added. Then, the control flow graphs (CFG) are built. Finally, we designed a GNN-based model for vulnerability detection. The experimental results show that the proposed method is effective for the vulnerability detection of smart contracts. Compared with the traditional approaches, it can get good results with fewer layers of the GCN model, which shows that the contract bytecode and GCN model are efficient in vulnerability detection.
comment: 8 pages, 8 figures
☆ RACCER: Towards Reachable and Certain Counterfactual Explanations for Reinforcement Learning
While reinforcement learning (RL) algorithms have been successfully applied to numerous tasks, their reliance on neural networks makes their behavior difficult to understand and trust. Counterfactual explanations are human-friendly explanations that offer users actionable advice on how to alter the model inputs to achieve the desired output from a black-box system. However, current approaches to generating counterfactuals in RL ignore the stochastic and sequential nature of RL tasks and can produce counterfactuals which are difficult to obtain or do not deliver the desired outcome. In this work, we propose RACCER, the first RL-specific approach to generating counterfactual explanations for the behaviour of RL agents. We first propose and implement a set of RL-specific counterfactual properties that ensure easily reachable counterfactuals with highly-probable desired outcomes. We use a heuristic tree search of agent's execution trajectories to find the most suitable counterfactuals based on the defined properties. We evaluate RACCER in two tasks as well as conduct a user study to show that RL-specific counterfactuals help users better understand agent's behavior compared to the current state-of-the-art approaches.
comment: 16 pages, 3 figures
☆ Grasping Student: semi-supervised learning for robotic manipulation
Gathering real-world data from the robot quickly becomes a bottleneck when constructing a robot learning system for grasping. In this work, we design a semi-supervised grasping system that, on top of a small sample of robot experience, takes advantage of images of products to be picked, which are collected without any interactions with the robot. We validate our findings both in the simulation and in the real world. In the regime of a small number of robot training samples, taking advantage of the unlabeled data allows us to achieve performance at the level of 10-fold bigger dataset size used by the baseline. The code and datasets used in the paper will be released at https://github.com/nomagiclab/grasping-student.
☆ Nonlinear Kalman Filtering with Reparametrization Gradients
We introduce a novel nonlinear Kalman filter that utilizes reparametrization gradients. The widely used parametric approximation is based on a jointly Gaussian assumption of the state-space model, which is in turn equivalent to minimizing an approximation to the Kullback-Leibler divergence. It is possible to obtain better approximations using the alpha divergence, but the resulting problem is substantially more complex. In this paper, we introduce an alternate formulation based on an energy function, which can be optimized instead of the alpha divergence. The optimization can be carried out using reparametrization gradients, a technique that has recently been utilized in a number of deep learning models.
☆ Loss-Curvature Matching for Dataset Selection and Condensation AISTATS
Training neural networks on a large dataset requires substantial computational costs. Dataset reduction selects or synthesizes data instances based on the large dataset, while minimizing the degradation in generalization performance from the full dataset. Existing methods utilize the neural network during the dataset reduction procedure, so the model parameter becomes important factor in preserving the performance after reduction. By depending upon the importance of parameters, this paper introduces a new reduction objective, coined LCMat, which Matches the Loss Curvatures of the original dataset and reduced dataset over the model parameter space, more than the parameter point. This new objective induces a better adaptation of the reduced dataset on the perturbed parameter region than the exact point matching. Particularly, we identify the worst case of the loss curvature gap from the local parameter region, and we derive the implementable upper bound of such worst-case with theoretical analyses. Our experiments on both coreset selection and condensation benchmarks illustrate that LCMat shows better generalization performances than existing baselines.
comment: 26th International Conference on Artificial Intelligence and Statistics (AISTATS)
☆ MKL-$L_{0/1}$-SVM
We formulate the Multiple Kernel Learning (abbreviated as MKL) problem for the support vector machine with the infamous $(0,1)$-loss function. Some first-order optimality conditions are given, which could be readily exploited to develop fast numerical solvers e.g., of the ADMM type.
comment: 4 pages, 2 figures. Submitted to the 3rd Chinese Conference on Predictive Control and Intelligent Decision (CPCID)
☆ HyT-NAS: Hybrid Transformers Neural Architecture Search for Edge Devices
Vision Transformers have enabled recent attention-based Deep Learning (DL) architectures to achieve remarkable results in Computer Vision (CV) tasks. However, due to the extensive computational resources required, these architectures are rarely implemented on resource-constrained platforms. Current research investigates hybrid handcrafted convolution-based and attention-based models for CV tasks such as image classification and object detection. In this paper, we propose HyT-NAS, an efficient Hardware-aware Neural Architecture Search (HW-NAS) including hybrid architectures targeting vision tasks on tiny devices. HyT-NAS improves state-of-the-art HW-NAS by enriching the search space and enhancing the search strategy as well as the performance predictors. Our experiments show that HyT-NAS achieves a similar hypervolume with less than ~5x training evaluations. Our resulting architecture outperforms MLPerf MobileNetV1 by 6.3% accuracy improvement with 3.5x less number of parameters on Visual Wake Words.
comment: CODAI 2022 Workshop - Embedded System Week (ESWeek)
☆ Learning Hybrid Interpretable Models: Theory, Taxonomy, and Methods
A hybrid model involves the cooperation of an interpretable model and a complex black box. At inference, any input of the hybrid model is assigned to either its interpretable or complex component based on a gating mechanism. The advantages of such models over classical ones are two-fold: 1) They grant users precise control over the level of transparency of the system and 2) They can potentially perform better than a standalone black box since redirecting some of the inputs to an interpretable model implicitly acts as regularization. Still, despite their high potential, hybrid models remain under-studied in the interpretability/explainability literature. In this paper, we remedy this fact by presenting a thorough investigation of such models from three perspectives: Theory, Taxonomy, and Methods. First, we explore the theory behind the generalization of hybrid models from the Probably-Approximately-Correct (PAC) perspective. A consequence of our PAC guarantee is the existence of a sweet spot for the optimal transparency of the system. When such a sweet spot is attained, a hybrid model can potentially perform better than a standalone black box. Secondly, we provide a general taxonomy for the different ways of training hybrid models: the Post-Black-Box and Pre-Black-Box paradigms. These approaches differ in the order in which the interpretable and complex components are trained. We show where the state-of-the-art hybrid models Hybrid-Rule-Set and Companion-Rule-List fall in this taxonomy. Thirdly, we implement the two paradigms in a single method: HybridCORELS, which extends the CORELS algorithm to hybrid modeling. By leveraging CORELS, HybridCORELS provides a certificate of optimality of its interpretable component and precise control over transparency. We finally show empirically that HybridCORELS is competitive with existing hybrid models, and performs just as well as a standalone black box (or even better) while being partly transparent.
☆ A comparison of rational and neural network based approximations
Rational and neural network based approximations are efficient tools in modern approximation. These approaches are able to produce accurate approximations to nonsmooth and non-Lipschitz functions, including multivariate domain functions. In this paper we compare the efficiency of function approximation using rational approximation, neural network and their combinations. It was found that rational approximation is superior to neural network based approaches with the same number of decision variables. Our numerical experiments demonstrate the efficiency of rational approximation, even when the number of approximation parameters (that is, the dimension of the corresponding optimisation problems) is small. Another important contribution of this paper lies in the improvement of rational approximation algorithms. Namely, the optimisation based algorithms for rational approximation can be adjusted to in such a way that the conditioning number of the constraint matrices are controlled. This simple adjustment enables us to work with high dimension optimisation problems and improve the design of the neural network. The main strength of neural networks is in their ability to handle models with a large number of variables: complex models are decomposed in several simple optimisation problems. Therefore the the large number of decision variables is in the nature of neural networks.
comment: 36 pages
☆ A Message Passing Perspective on Learning Dynamics of Contrastive Learning ICLR 2023
In recent years, contrastive learning achieves impressive results on self-supervised visual representation learning, but there still lacks a rigorous understanding of its learning dynamics. In this paper, we show that if we cast a contrastive objective equivalently into the feature space, then its learning dynamics admits an interpretable form. Specifically, we show that its gradient descent corresponds to a specific message passing scheme on the corresponding augmentation graph. Based on this perspective, we theoretically characterize how contrastive learning gradually learns discriminative features with the alignment update and the uniformity update. Meanwhile, this perspective also establishes an intriguing connection between contrastive learning and Message Passing Graph Neural Networks (MP-GNNs). This connection not only provides a unified understanding of many techniques independently developed in each community, but also enables us to borrow techniques from MP-GNNs to design new contrastive learning variants, such as graph attention, graph rewiring, jumpy knowledge techniques, etc. We believe that our message passing perspective not only provides a new theoretical understanding of contrastive learning dynamics, but also bridges the two seemingly independent areas together, which could inspire more interleaving studies to benefit from each other. The code is available at https://github.com/PKU-ML/Message-Passing-Contrastive-Learning.
comment: ICLR 2023
☆ FUSQA: Fetal Ultrasound Segmentation Quality Assessment
Deep learning models have been effective for various fetal ultrasound segmentation tasks. However, generalization to new unseen data has raised questions about their effectiveness for clinical adoption. Normally, a transition to new unseen data requires time-consuming and costly quality assurance processes to validate the segmentation performance post-transition. Segmentation quality assessment efforts have focused on natural images, where the problem has been typically formulated as a dice score regression task. In this paper, we propose a simplified Fetal Ultrasound Segmentation Quality Assessment (FUSQA) model to tackle the segmentation quality assessment when no masks exist to compare with. We formulate the segmentation quality assessment process as an automated classification task to distinguish between good and poor-quality segmentation masks for more accurate gestational age estimation. We validate the performance of our proposed approach on two datasets we collect from two hospitals using different ultrasound machines. We compare different architectures, with our best-performing architecture achieving over 90% classification accuracy on distinguishing between good and poor-quality segmentation masks from an unseen dataset. Additionally, there was only a 1.45-day difference between the gestational age reported by doctors and estimated based on CRL measurements using well-segmented masks. On the other hand, this difference increased and reached up to 7.73 days when we calculated CRL from the poorly segmented masks. As a result, AI-based approaches can potentially aid fetal ultrasound segmentation quality assessment and might detect poor segmentation in real-time screening in the future.
comment: 13 pages, 3 figures, 3 tables
☆ Inference on Optimal Dynamic Policies via Softmax Approximation
Estimating optimal dynamic policies from offline data is a fundamental problem in dynamic decision making. In the context of causal inference, the problem is known as estimating the optimal dynamic treatment regime. Even though there exists a plethora of methods for estimation, constructing confidence intervals for the value of the optimal regime and structural parameters associated with it is inherently harder, as it involves non-linear and non-differentiable functionals of un-known quantities that need to be estimated. Prior work resorted to sub-sample approaches that can deteriorate the quality of the estimate. We show that a simple soft-max approximation to the optimal treatment regime, for an appropriately fast growing temperature parameter, can achieve valid inference on the truly optimal regime. We illustrate our result for a two-period optimal dynamic regime, though our approach should directly extend to the finite horizon case. Our work combines techniques from semi-parametric inference and $g$-estimation, together with an appropriate triangular array central limit theorem, as well as a novel analysis of the asymptotic influence and asymptotic bias of softmax approximations.
☆ The Lie-Group Bayesian Learning Rule AISTATS 2023
The Bayesian Learning Rule provides a framework for generic algorithm design but can be difficult to use for three reasons. First, it requires a specific parameterization of exponential family. Second, it uses gradients which can be difficult to compute. Third, its update may not always stay on the manifold. We address these difficulties by proposing an extension based on Lie-groups where posteriors are parametrized through transformations of an arbitrary base distribution and updated via the group's exponential map. This simplifies all three difficulties for many cases, providing flexible parametrizations through group's action, simple gradient computation through reparameterization, and updates that always stay on the manifold. We use the new learning rule to derive a new algorithm for deep learning with desirable biologically-plausible attributes to learn sparse features. Our work opens a new frontier for the design of new algorithms by exploiting Lie-group structures.
comment: AISTATS 2023
☆ A Deep-Learning-Based Neural Decoding Framework for Emotional Brain-Computer Interfaces
Reading emotions precisely from segments of neural activity is crucial for the development of emotional brain-computer interfaces. Among all neural decoding algorithms, deep learning (DL) holds the potential to become the most promising one, yet progress has been limited in recent years. One possible reason is that the efficacy of DL strongly relies on training samples, yet the neural data used for training are often from non-human primates and mixed with plenty of noise, which in turn mislead the training of DL models. Given it is difficult to accurately determine animals' emotions from humans' perspective, we assume the dominant noise in neural data representing different emotions is the labeling error. Here, we report the development and application of a neural decoding framework called Emo-Net that consists of a confidence learning (CL) component and a DL component. The framework is fully data-driven and is capable of decoding emotions from multiple datasets obtained from behaving monkeys. In addition to improving the decoding ability, Emo-Net significantly improves the performance of the base DL models, making emotion recognition in animal models possible. In summary, this framework may inspire novel understandings of the neural basis of emotion and drive the realization of close-loop emotional brain-computer interfaces.
comment: 22 pages, 4 figures
☆ Policy Mirror Descent Inherently Explores Action Space
Designing computationally efficient exploration strategies for on-policy first-order methods that attain optimal $\mathcal{O}(1/\epsilon^2)$ sample complexity remains open for solving Markov decision processes (MDP). This manuscript provides an answer to this question from a perspective of simplicity, by showing that whenever exploration over the state space is implied by the MDP structure, there seems to be little need for sophisticated exploration strategies. We revisit a stochastic policy gradient method, named stochastic policy mirror descent, applied to the infinite horizon, discounted MDP with finite state and action spaces. Accompanying SPMD we present two on-policy evaluation operators, both simply following the policy for trajectory collection with no explicit exploration, or any form of intervention. SPMD with the first evaluation operator, named value-based estimation, tailors to the Kullback-Leibler (KL) divergence. Provided the Markov chains on the state space of generated policies are uniformly mixing with non-diminishing minimal visitation measure, an $\tilde{\mathcal{O}}( 1 / \epsilon^2)$ sample complexity is obtained with a linear dependence on the size of the action space. SPMD with the second evaluation operator, named truncated on-policy Monte Carlo, attains an $\tilde{\mathcal{O}}(\mathcal{H}_{\mathcal{D}} / \epsilon^2)$ sample complexity, with the same assumption on the state chains of generated policies. We characterize $\mathcal{H}_{\mathcal{D}}$ as a divergence-dependent function of the effective horizon and the size of the action space, which leads to an exponential dependence of the latter two quantities for the KL divergence, and a polynomial dependence for the divergence induced by negative Tsallis entropy. These obtained sample complexities seem to be new among on-policy stochastic policy gradient methods without explicit explorations.
☆ Automatically Auditing Large Language Models via Discrete Optimization
Auditing large language models for unexpected behaviors is critical to preempt catastrophic deployments, yet remains challenging. In this work, we cast auditing as an optimization problem, where we automatically search for input-output pairs that match a desired target behavior. For example, we might aim to find a non-toxic input that starts with "Barack Obama" that a model maps to a toxic output. This optimization problem is difficult to solve as the set of feasible points is sparse, the space is discrete, and the language models we audit are non-linear and high-dimensional. To combat these challenges, we introduce a discrete optimization algorithm, ARCA, that jointly and efficiently optimizes over inputs and outputs. Our approach automatically uncovers derogatory completions about celebrities (e.g. "Barack Obama is a legalized unborn" -> "child murderer"), produces French inputs that complete to English outputs, and finds inputs that generate a specific name. Our work offers a promising new tool to uncover models' failure-modes before deployment.
☆ HappyMap: A Generalized Multi-calibration Method
Multi-calibration is a powerful and evolving concept originating in the field of algorithmic fairness. For a predictor $f$ that estimates the outcome $y$ given covariates $x$, and for a function class $\mathcal{C}$, multi-calibration requires that the predictor $f(x)$ and outcome $y$ are indistinguishable under the class of auditors in $\mathcal{C}$. Fairness is captured by incorporating demographic subgroups into the class of functions~$\mathcal{C}$. Recent work has shown that, by enriching the class $\mathcal{C}$ to incorporate appropriate propensity re-weighting functions, multi-calibration also yields target-independent learning, wherein a model trained on a source domain performs well on unseen, future, target domains(approximately) captured by the re-weightings. Formally, multi-calibration with respect to $\mathcal{C}$ bounds $\big|\mathbb{E}_{(x,y)\sim \mathcal{D}}[c(f(x),x)\cdot(f(x)-y)]\big|$ for all $c \in \mathcal{C}$. In this work, we view the term $(f(x)-y)$ as just one specific mapping, and explore the power of an enriched class of mappings. We propose \textit{HappyMap}, a generalization of multi-calibration, which yields a wide range of new applications, including a new fairness notion for uncertainty quantification (conformal prediction), a novel technique for conformal prediction under covariate shift, and a different approach to analyzing missing data, while also yielding a unified understanding of several existing seemingly disparate algorithmic fairness notions and target-independent learning approaches. We give a single \textit{HappyMap} meta-algorithm that captures all these results, together with a sufficiency condition for its success.
comment: Appeared at ITCS 2023 (submitted on Sept. 8th, 2022)
☆ Semantically Consistent Multi-view Representation Learning
In this work, we devote ourselves to the challenging task of Unsupervised Multi-view Representation Learning (UMRL), which requires learning a unified feature representation from multiple views in an unsupervised manner. Existing UMRL methods mainly concentrate on the learning process in the feature space while ignoring the valuable semantic information hidden in different views. To address this issue, we propose a novel Semantically Consistent Multi-view Representation Learning (SCMRL), which makes efforts to excavate underlying multi-view semantic consensus information and utilize the information to guide the unified feature representation learning. Specifically, SCMRL consists of a within-view reconstruction module and a unified feature representation learning module, which are elegantly integrated by the contrastive learning strategy to simultaneously align semantic labels of both view-specific feature representations and the learned unified feature representation. In this way, the consensus information in the semantic space can be effectively exploited to constrain the learning process of unified feature representation. Compared with several state-of-the-art algorithms, extensive experiments demonstrate its superiority.
comment: 19 pages, 4figures
☆ Dynamic Scenario Representation Learning for Motion Forecasting with Heterogeneous Graph Convolutional Recurrent Networks
Due to the complex and changing interactions in dynamic scenarios, motion forecasting is a challenging problem in autonomous driving. Most existing works exploit static road graphs to characterize scenarios and are limited in modeling evolving spatio-temporal dependencies in dynamic scenarios. In this paper, we resort to dynamic heterogeneous graphs to model the scenario. Various scenario components including vehicles (agents) and lanes, multi-type interactions, and their changes over time are jointly encoded. Furthermore, we design a novel heterogeneous graph convolutional recurrent network, aggregating diverse interaction information and capturing their evolution, to learn to exploit intrinsic spatio-temporal dependencies in dynamic graphs and obtain effective representations of dynamic scenarios. Finally, with a motion forecasting decoder, our model predicts realistic and multi-modal future trajectories of agents and outperforms state-of-the-art published works on several motion forecasting benchmarks.
☆ Does Synthetic Data Generation of LLMs Help Clinical Text Mining?
Recent advancements in large language models (LLMs) have led to the development of highly potent models like OpenAI's ChatGPT. These models have exhibited exceptional performance in a variety of tasks, such as question answering, essay composition, and code generation. However, their effectiveness in the healthcare sector remains uncertain. In this study, we seek to investigate the potential of ChatGPT to aid in clinical text mining by examining its ability to extract structured information from unstructured healthcare texts, with a focus on biological named entity recognition and relation extraction. However, our preliminary results indicate that employing ChatGPT directly for these tasks resulted in poor performance and raised privacy concerns associated with uploading patients' information to the ChatGPT API. To overcome these limitations, we propose a new training paradigm that involves generating a vast quantity of high-quality synthetic data with labels utilizing ChatGPT and fine-tuning a local model for the downstream task. Our method has resulted in significant improvements in the performance of downstream tasks, improving the F1-score from 23.37% to 63.99% for the named entity recognition task and from 75.86% to 83.59% for the relation extraction task. Furthermore, generating data using ChatGPT can significantly reduce the time and effort required for data collection and labeling, as well as mitigate data privacy concerns. In summary, the proposed framework presents a promising solution to enhance the applicability of LLM models to clinical text mining.
comment: 10 pages, 8 tables, 4 figures
☆ Soft Actor-Critic Algorithm with Truly Inequality Constraint
Soft actor-critic (SAC) in reinforcement learning is expected to be one of the next-generation robot control schemes. Its ability to maximize policy entropy would make a robotic controller robust to noise and perturbation, which is useful for real-world robot applications. However, the priority of maximizing the policy entropy is automatically tuned in the current implementation, the rule of which can be interpreted as one for equality constraint, binding the policy entropy into its specified target value. The current SAC is therefore no longer maximize the policy entropy, contrary to our expectation. To resolve this issue in SAC, this paper improves its implementation with a slack variable for appropriately handling the inequality constraint to maximize the policy entropy. In Mujoco and Pybullet simulators, the modified SAC achieved the higher robustness and the more stable learning than before while regularizing the norm of action. In addition, a real-robot variable impedance task was demonstrated for showing the applicability of the modified SAC to real-world robot control.
comment: 8 pages, 7 figures
☆ Federated Learning via Variational Bayesian Inference: Personalization, Sparsity and Clustering
Federated learning (FL) is a promising framework that models distributed machine learning while protecting the privacy of clients. However, FL suffers performance degradation from heterogeneous and limited data. To alleviate the degradation, we present a novel personalized Bayesian FL approach named pFedBayes. By using the trained global distribution from the server as the prior distribution of each client, each client adjusts its own distribution by minimizing the sum of the reconstruction error over its personalized data and the KL divergence with the downloaded global distribution. Then, we propose a sparse personalized Bayesian FL approach named sFedBayes. To overcome the extreme heterogeneity in non-i.i.d. data, we propose a clustered Bayesian FL model named cFedbayes by learning different prior distributions for different clients. Theoretical analysis gives the generalization error bound of three approaches and shows that the generalization error convergence rates of the proposed approaches achieve minimax optimality up to a logarithmic factor. Moreover, the analysis presents that cFedbayes has a tighter generalization error rate than pFedBayes. Numerous experiments are provided to demonstrate that the proposed approaches have better performance than other advanced personalized methods on private models in the presence of heterogeneous and limited data.
comment: 17 pages, 19 figures
☆ Privacy-preserving and Uncertainty-aware Federated Trajectory Prediction for Connected Autonomous Vehicles
Deep learning is the method of choice for trajectory prediction for autonomous vehicles. Unfortunately, its data-hungry nature implicitly requires the availability of sufficiently rich and high-quality centralized datasets, which easily leads to privacy leakage. Besides, uncertainty-awareness becomes increasingly important for safety-crucial cyber physical systems whose prediction module heavily relies on machine learning tools. In this paper, we relax the data collection requirement and enhance uncertainty-awareness by using Federated Learning on Connected Autonomous Vehicles with an uncertainty-aware global objective. We name our algorithm as FLTP. We further introduce ALFLTP which boosts FLTP via using active learning techniques in adaptatively selecting participating clients. We consider both negative log-likelihood (NLL) and aleatoric uncertainty (AU) as client selection metrics. Experiments on Argoverse dataset show that FLTP significantly outperforms the model trained on local data. In addition, ALFLTP-AU converges faster in training regression loss and performs better in terms of NLL, minADE and MR than FLTP in most rounds, and has more stable round-wise performance than ALFLTP-NLL.
☆ Learning the Finer Things: Bayesian Structure Learning at the Instantiation Level
Successful machine learning methods require a trade-off between memorization and generalization. Too much memorization and the model cannot generalize to unobserved examples. Too much over-generalization and we risk under-fitting the data. While we commonly measure their performance through cross validation and accuracy metrics, how should these algorithms cope in domains that are extremely under-determined where accuracy is always unsatisfactory? We present a novel probabilistic graphical model structure learning approach that can learn, generalize and explain in these elusive domains by operating at the random variable instantiation level. Using Minimum Description Length (MDL) analysis, we propose a new decomposition of the learning problem over all training exemplars, fusing together minimal entropy inferences to construct a final knowledge base. By leveraging Bayesian Knowledge Bases (BKBs), a framework that operates at the instantiation level and inherently subsumes Bayesian Networks (BNs), we develop both a theoretical MDL score and associated structure learning algorithm that demonstrates significant improvements over learned BNs on 40 benchmark datasets. Further, our algorithm incorporates recent off-the-shelf DAG learning techniques enabling tractable results even on large problems. We then demonstrate the utility of our approach in a significantly under-determined domain by learning gene regulatory networks on breast cancer gene mutational data available from The Cancer Genome Atlas (TCGA).
☆ Provable Pathways: Learning Multiple Tasks over Multiple Paths
Constructing useful representations across a large number of tasks is a key requirement for sample-efficient intelligent systems. A traditional idea in multitask learning (MTL) is building a shared representation across tasks which can then be adapted to new tasks by tuning last layers. A desirable refinement of using a shared one-fits-all representation is to construct task-specific representations. To this end, recent PathNet/muNet architectures represent individual tasks as pathways within a larger supernet. The subnetworks induced by pathways can be viewed as task-specific representations that are composition of modules within supernet's computation graph. This work explores the pathways proposal from the lens of statistical learning: We first develop novel generalization bounds for empirical risk minimization problems learning multiple tasks over multiple paths (Multipath MTL). In conjunction, we formalize the benefits of resulting multipath representation when adapting to new downstream tasks. Our bounds are expressed in terms of Gaussian complexity, lead to tangible guarantees for the class of linear representations, and provide novel insights into the quality and benefits of a multipath representation. When computation graph is a tree, Multipath MTL hierarchically clusters the tasks and builds cluster-specific representations. We provide further discussion and experiments for hierarchical MTL and rigorously identify the conditions under which Multipath MTL is provably superior to traditional MTL approaches with shallow supernets.
☆ QuickSRNet: Plain Single-Image Super-Resolution Architecture for Faster Inference on Mobile Platforms
In this work, we present QuickSRNet, an efficient super-resolution architecture for real-time applications on mobile platforms. Super-resolution clarifies, sharpens, and upscales an image to higher resolution. Applications such as gaming and video playback along with the ever-improving display capabilities of TVs, smartphones, and VR headsets are driving the need for efficient upscaling solutions. While existing deep learning-based super-resolution approaches achieve impressive results in terms of visual quality, enabling real-time DL-based super-resolution on mobile devices with compute, thermal, and power constraints is challenging. To address these challenges, we propose QuickSRNet, a simple yet effective architecture that provides better accuracy-to-latency trade-offs than existing neural architectures for single-image super resolution. We present training tricks to speed up existing residual-based super-resolution architectures while maintaining robustness to quantization. Our proposed architecture produces 1080p outputs via 2x upscaling in 2.2 ms on a modern smartphone, making it ideal for high-fps real-time applications.
comment: 16 pages
☆ Unbiased Learning to Rank with Biased Continuous Feedback
It is a well-known challenge to learn an unbiased ranker with biased feedback. Unbiased learning-to-rank(LTR) algorithms, which are verified to model the relative relevance accurately based on noisy feedback, are appealing candidates and have already been applied in many applications with single categorical labels, such as user click signals. Nevertheless, the existing unbiased LTR methods cannot properly handle continuous feedback, which are essential for many industrial applications, such as content recommender systems. To provide personalized high-quality recommendation results, recommender systems need model both categorical and continuous biased feedback, such as click and dwell time. Accordingly, we design a novel unbiased LTR algorithm to tackle the challenges, which innovatively models position bias in the pairwise fashion and introduces the pairwise trust bias to separate the position bias, trust bias, and user relevance explicitly and can work for both continuous and categorical feedback. Experiment results on public benchmark datasets and internal live traffic of a large-scale recommender system at Tencent News show superior results for continuous labels and also competitive performance for categorical labels of the proposed method.
comment: 10 pages. arXiv admin note: substantial text overlap with arXiv:2111.12929
☆ Preference-Aware Delivery Planning for Last-Mile Logistics AAMAS-23
Optimizing delivery routes for last-mile logistics service is challenging and has attracted the attention of many researchers. These problems are usually modeled and solved as variants of vehicle routing problems (VRPs) with challenging real-world constraints (e.g., time windows, precedence). However, despite many decades of solid research on solving these VRP instances, we still see significant gaps between optimized routes and the routes that are actually preferred by the practitioners. Most of these gaps are due to the difference between what's being optimized, and what the practitioners actually care about, which is hard to be defined exactly in many instances. In this paper, we propose a novel hierarchical route optimizer with learnable parameters that combines the strength of both the optimization and machine learning approaches. Our hierarchical router first solves a zone-level Traveling Salesman Problem with learnable weights on various zone-level features; with the zone visit sequence fixed, we then solve the stop-level vehicle routing problem as a Shortest Hamiltonian Path problem. The Bayesian optimization approach is then introduced to allow us to adjust the weights to be assigned to different zone features used in solving the zone-level Traveling Salesman Problem. By using a real-world delivery dataset provided by the Amazon Last Mile Routing Research Challenge, we demonstrate the importance of having both the optimization and the machine learning components. We also demonstrate how we can use route-related features to identify instances that we might have difficulty with. This paves ways to further research on how we can tackle these difficult instances.
comment: Accepted to the 22nd International Conference on Autonomous Agents and Multiagent Systems (AAMAS-23)
☆ The Novel Adaptive Fractional Order Gradient Decent Algorithms Design via Robust Control
The vanilla fractional order gradient descent may oscillatively converge to a region around the global minimum instead of converging to the exact minimum point, or even diverge, in the case where the objective function is strongly convex. To address this problem, a novel adaptive fractional order gradient descent (AFOGD) method and a novel adaptive fractional order accelerated gradient descent (AFOAGD) method are proposed in this paper. Inspired by the quadratic constraints and Lyapunov stability analysis from robust control theory, we establish a linear matrix inequality to analyse the convergence of our proposed algorithms. We prove that the proposed algorithms can achieve R-linear convergence when the objective function is $\textbf{L-}$smooth and $\textbf{m-}$strongly-convex. Several numerical simulations are demonstrated to verify the effectiveness and superiority of our proposed algorithms.
comment: 8pages,5 figures
☆ Using Memory-Based Learning to Solve Tasks with State-Action Constraints
Tasks where the set of possible actions depend discontinuously on the state pose a significant challenge for current reinforcement learning algorithms. For example, a locked door must be first unlocked, and then the handle turned before the door can be opened. The sequential nature of these tasks makes obtaining final rewards difficult, and transferring information between task variants using continuous learned values such as weights rather than discrete symbols can be inefficient. Our key insight is that agents that act and think symbolically are often more effective in dealing with these tasks. We propose a memory-based learning approach that leverages the symbolic nature of constraints and temporal ordering of actions in these tasks to quickly acquire and transfer high-level information. We evaluate the performance of memory-based learning on both real and simulated tasks with approximately discontinuous constraints between states and actions, and show our method learns to solve these tasks an order of magnitude faster than both model-based and model-free deep reinforcement learning methods.
comment: 8 pages, 3 figures, accepted to the International Conference on Robotics and Automation 2023
☆ Learning Environment-Aware Control Barrier Functions for Safe and Feasible Multi-Robot Navigation
Control Barrier Functions (CBFs) have been applied to provide safety guarantees for robot navigation. Traditional approaches consider fixed CBFs during navigation and hand-tune the underlying parameters apriori. Such approaches are inefficient and vulnerable to changes in the environment. The goal of this paper is to learn CBFs for multi-robot navigation based on what robots perceive about their environment. In order to guarantee the feasibility of the navigation task, while ensuring robot safety, we pursue a trade-off between conservativeness and aggressiveness in robot behavior by defining dynamic environment-aware CBF constraints. Since the explicit relationship between CBF constraints and navigation performance is challenging to model, we leverage reinforcement learning to learn time-varying CBFs in a model-free manner. We parameterize the CBF policy with graph neural networks (GNNs), and design GNNs that are translation invariant and permutation equivariant, to synthesize decentralized policies that generalize across environments. The proposed approach maintains safety guarantees (due to the underlying CBFs), while optimizing navigation performance (due to the reward-based learning). We perform simulations that compare the proposed approach with fixed CBFs tuned by exhaustive grid-search. The results show that environment-aware CBFs are capable of adapting to robot movements and obstacle changes, yielding improved navigation performance and robust generalization.
☆ Optimal Sparse Recovery with Decision Stumps AAAI 2023
Decision trees are widely used for their low computational cost, good predictive performance, and ability to assess the importance of features. Though often used in practice for feature selection, the theoretical guarantees of these methods are not well understood. We here obtain a tight finite sample bound for the feature selection problem in linear regression using single-depth decision trees. We examine the statistical properties of these "decision stumps" for the recovery of the $s$ active features from $p$ total features, where $s \ll p$. Our analysis provides tight sample performance guarantees on high-dimensional sparse systems which align with the finite sample bound of $O(s \log p)$ as obtained by Lasso, improving upon previous bounds for both the median and optimal splitting criteria. Our results extend to the non-linear regime as well as arbitrary sub-Gaussian distributions, demonstrating that tree based methods attain strong feature selection properties under a wide variety of settings and further shedding light on the success of these methods in practice. As a byproduct of our analysis, we show that we can provably guarantee recovery even when the number of active features $s$ is unknown. We further validate our theoretical results and proof methodology using computational experiments.
comment: Accepted to AAAI 2023
☆ ATM Fraud Detection using Streaming Data Analytics
Gaining the trust and confidence of customers is the essence of the growth and success of financial institutions and organizations. Of late, the financial industry is significantly impacted by numerous instances of fraudulent activities. Further, owing to the generation of large voluminous datasets, it is highly essential that underlying framework is scalable and meet real time needs. To address this issue, in the study, we proposed ATM fraud detection in static and streaming contexts respectively. In the static context, we investigated a parallel and scalable machine learning algorithms for ATM fraud detection that is built on Spark and trained with a variety of machine learning (ML) models including Naive Bayes (NB), Logistic Regression (LR), Support Vector Machine (SVM), Decision Tree (DT), Random Forest (RF), Gradient Boosting Tree (GBT), and Multi-layer perceptron (MLP). We also employed several balancing techniques like Synthetic Minority Oversampling Technique (SMOTE) and its variants, Generative Adversarial Networks (GAN), to address the rarity in the dataset. In addition, we proposed a streaming based ATM fraud detection in the streaming context. Our sliding window based method collects ATM transactions that are performed within a specified time interval and then utilizes to train several ML models, including NB, RF, DT, and K-Nearest Neighbour (KNN). We selected these models based on their less model complexity and quicker response time. In both contexts, RF turned out to be the best model. RF obtained the best mean AUC of 0.975 in the static context and mean AUC of 0.910 in the streaming context. RF is also empirically proven to be statistically significant than the next-best performing models.
comment: 25 pages, 15 figures, 10 tables. arXiv admin note: text overlap with arXiv:2211.10595
☆ On the Benefits of Biophysical Synapses
The approximation capability of ANNs and their RNN instantiations, is strongly correlated with the number of parameters packed into these networks. However, the complexity barrier for human understanding, is arguably related to the number of neurons and synapses in the networks, and to the associated nonlinear transformations. In this paper we show that the use of biophysical synapses, as found in LTCs, have two main benefits. First, they allow to pack more parameters for a given number of neurons and synapses. Second, they allow to formulate the nonlinear-network transformation, as a linear system with state-dependent coefficients. Both increase interpretability, as for a given task, they allow to learn a system linear in its input features, that is smaller in size compared to the state of the art. We substantiate the above claims on various time-series prediction tasks, but we believe that our results are applicable to any feedforward or recurrent ANN.
☆ A Study of Variable-Role-based Feature Enrichment in Neural Models of Code
Although deep neural models substantially reduce the overhead of feature engineering, the features readily available in the inputs might significantly impact training cost and the performance of the models. In this paper, we explore the impact of an unsuperivsed feature enrichment approach based on variable roles on the performance of neural models of code. The notion of variable roles (as introduced in the works of Sajaniemi et al. [Refs. 1,2]) has been found to help students' abilities in programming. In this paper, we investigate if this notion would improve the performance of neural models of code. To the best of our knowledge, this is the first work to investigate how Sajaniemi et al.'s concept of variable roles can affect neural models of code. In particular, we enrich a source code dataset by adding the role of individual variables in the dataset programs, and thereby conduct a study on the impact of variable role enrichment in training the Code2Seq model. In addition, we shed light on some challenges and opportunities in feature enrichment for neural code intelligence models.
comment: Accepted in the 1st International Workshop on Interpretability and Robustness in Neural Software Engineering (InteNSE'23), Co-located with ICSE
☆ Multimodal Multi-User Surface Recognition with the Kernel Two-Sample Test
Machine learning and deep learning have been used extensively to classify physical surfaces through images and time-series contact data. However, these methods rely on human expertise and entail the time-consuming processes of data and parameter tuning. To overcome these challenges, we propose an easily implemented framework that can directly handle heterogeneous data sources for classification tasks. Our data-versus-data approach automatically quantifies distinctive differences in distributions in a high-dimensional space via kernel two-sample testing between two sets extracted from multimodal data (e.g., images, sounds, haptic signals). We demonstrate the effectiveness of our technique by benchmarking against expertly engineered classifiers for visual-audio-haptic surface recognition due to the industrial relevance, difficulty, and competitive baselines of this application; ablation studies confirm the utility of key components of our pipeline. As shown in our open-source code, we achieve 97.2% accuracy on a standard multi-user dataset with 108 surface classes, outperforming the state-of-the-art machine-learning algorithm by 6% on a more difficult version of the task. The fact that our classifier obtains this performance with minimal data processing in the standard algorithm setting reinforces the powerful nature of kernel methods for learning to recognize complex patterns.
☆ Embodied Active Learning of Relational State Abstractions for Bilevel Planning
State abstraction is an effective technique for planning in robotics environments with continuous states and actions, long task horizons, and sparse feedback. In object-oriented environments, predicates are a particularly useful form of state abstraction because of their compatibility with symbolic planners and their capacity for relational generalization. However, to plan with predicates, the agent must be able to interpret them in continuous environment states (i.e., ground the symbols). Manually programming predicate interpretations can be difficult, so we would instead like to learn them from data. We propose an embodied active learning paradigm where the agent learns predicate interpretations through online interaction with an expert. For example, after taking actions in a block stacking environment, the agent may ask the expert: "Is On(block1, block2) true?" From this experience, the agent learns to plan: it learns neural predicate interpretations, symbolic planning operators, and neural samplers that can be used for bilevel planning. During exploration, the agent plans to learn: it uses its current models to select actions towards generating informative expert queries. We learn predicate interpretations as ensembles of neural networks and use their entropy to measure the informativeness of potential queries. We evaluate this approach in three robotic environments and find that it consistently outperforms six baselines while exhibiting sample efficiency in two key metrics: number of environment interactions, and number of queries to the expert. Code: https://tinyurl.com/active-predicates
☆ Reverse Engineering Breast MRIs: Predicting Acquisition Parameters Directly from Images
The image acquisition parameters (IAPs) used to create MRI scans are central to defining the appearance of the images. Deep learning models trained on data acquired using certain parameters might not generalize well to images acquired with different parameters. Being able to recover such parameters directly from an image could help determine whether a deep learning model is applicable, and could assist with data harmonization and/or domain adaptation. Here, we introduce a neural network model that can predict many complex IAPs used to generate an MR image with high accuracy solely using the image, with a single forward pass. These predicted parameters include field strength, echo and repetition times, acquisition matrix, scanner model, scan options, and others. Even challenging parameters such as contrast agent type can be predicted with good accuracy. We perform a variety of experiments and analyses of our model's ability to predict IAPs on many MRI scans of new patients, and demonstrate its usage in a realistic application. Predicting IAPs from the images is an important step toward better understanding the relationship between image appearance and IAPs. This in turn will advance the understanding of many concepts related to the generalizability of neural network models on medical images, including domain shift, domain adaptation, and data harmonization.
comment: Paper accepted at MIDL 2023. Code available at https://github.com/mazurowski-lab/MRI-IAP-prediction
☆ Baldur: Whole-Proof Generation and Repair with Large Language Models
Formally verifying software properties is a highly desirable but labor-intensive task. Recent work has developed methods to automate formal verification using proof assistants, such as Coq and Isabelle/HOL, e.g., by training a model to predict one proof step at a time, and using that model to search through the space of possible proofs. This paper introduces a new method to automate formal verification: We use large language models, trained on natural language text and code and fine-tuned on proofs, to generate whole proofs for theorems at once, rather than one step at a time. We combine this proof generation model with a fine-tuned repair model to repair generated proofs, further increasing proving power. As its main contributions, this paper demonstrates for the first time that: (1) Whole-proof generation using transformers is possible and is as effective as search-based techniques without requiring costly search. (2) Giving the learned model additional context, such as a prior failed proof attempt and the ensuing error message, results in proof repair and further improves automated proof generation. (3) We establish a new state of the art for fully automated proof synthesis. We reify our method in a prototype, Baldur, and evaluate it on a benchmark of 6,336 Isabelle/HOL theorems and their proofs. In addition to empirically showing the effectiveness of whole-proof generation, repair, and added context, we show that Baldur improves on the state-of-the-art tool, Thor, by automatically generating proofs for an additional 8.7% of the theorems. Together, Baldur and Thor can prove 65.7% of the theorems fully automatically. This paper paves the way for new research into using large language models for automating formal verification.
☆ Model-Agnostic Federated Learning
Since its debut in 2016, Federated Learning (FL) has been tied to the inner workings of Deep Neural Networks (DNNs). On the one hand, this allowed its development and widespread use as DNNs proliferated. On the other hand, it neglected all those scenarios in which using DNNs is not possible or advantageous. The fact that most current FL frameworks only allow training DNNs reinforces this problem. To address the lack of FL solutions for non-DNN-based use cases, we propose MAFL (Model-Agnostic Federated Learning). MAFL marries a model-agnostic FL algorithm, AdaBoost.F, with an open industry-grade FL framework: Intel OpenFL. MAFL is the first FL system not tied to any specific type of machine learning model, allowing exploration of FL scenarios beyond DNNs and trees. We test MAFL from multiple points of view, assessing its correctness, flexibility and scaling properties up to 64 nodes. We optimised the base software achieving a 5.5x speedup on a standard FL scenario. MAFL is compatible with x86-64, ARM-v8, Power and RISC-V.
☆ You Only Crash Once: Improved Object Detection for Real-Time, Sim-to-Real Hazardous Terrain Detection and Classification for Autonomous Planetary Landings
The detection of hazardous terrain during the planetary landing of spacecraft plays a critical role in assuring vehicle safety and mission success. A cheap and effective way of detecting hazardous terrain is through the use of visual cameras, which ensure operational ability from atmospheric entry through touchdown. Plagued by resource constraints and limited computational power, traditional techniques for visual hazardous terrain detection focus on template matching and registration to pre-built hazard maps. Although successful on previous missions, this approach is restricted to the specificity of the templates and limited by the fidelity of the underlying hazard map, which both require extensive pre-flight cost and effort to obtain and develop. Terrestrial systems that perform a similar task in applications such as autonomous driving utilize state-of-the-art deep learning techniques to successfully localize and classify navigation hazards. Advancements in spacecraft co-processors aimed at accelerating deep learning inference enable the application of these methods in space for the first time. In this work, we introduce You Only Crash Once (YOCO), a deep learning-based visual hazardous terrain detection and classification technique for autonomous spacecraft planetary landings. Through the use of unsupervised domain adaptation we tailor YOCO for training by simulation, removing the need for real-world annotated data and expensive mission surveying phases. We further improve the transfer of representative terrain knowledge between simulation and the real world through visual similarity clustering. We demonstrate the utility of YOCO through a series of terrestrial and extraterrestrial simulation-to-real experiments and show substantial improvements toward the ability to both detect and accurately classify instances of planetary terrain.
comment: To be published in proceedings of AAS/AIAA Astrodynamics Specialist Conference 2022
☆ Memory-adaptive Depth-wise Heterogenous Federated Learning
Federated learning is a promising paradigm that allows multiple clients to collaboratively train a model without sharing the local data. However, the presence of heterogeneous devices in federated learning, such as mobile phones and IoT devices with varying memory capabilities, would limit the scale and hence the performance of the model could be trained. The mainstream approaches to address memory limitations focus on width-slimming techniques, where different clients train subnetworks with reduced widths locally and then the server aggregates the subnetworks. The global model produced from these methods suffers from performance degradation due to the negative impact of the actions taken to handle the varying subnetwork widths in the aggregation phase. In this paper, we introduce a memory-adaptive depth-wise learning solution in FL called FeDepth, which adaptively decomposes the full model into blocks according to the memory budgets of each client and trains blocks sequentially to obtain a full inference model. Our method outperforms state-of-the-art approaches, achieving 5% and more than 10% improvements in top-1 accuracy on CIFAR-10 and CIFAR-100, respectively. We also demonstrate the effectiveness of depth-wise fine-tuning on ViT. Our findings highlight the importance of memory-aware techniques for federated learning with heterogeneous devices and the success of depth-wise training strategy in improving the global model's performance.
☆ DeepGD: A Multi-Objective Black-Box Test Selection Approach for Deep Neural Networks
Deep neural networks (DNNs) are widely used in various application domains such as image processing, speech recognition, and natural language processing. However, testing DNN models may be challenging due to the complexity and size of their input domain. Particularly, testing DNN models often requires generating or exploring large unlabeled datasets. In practice, DNN test oracles, which identify the correct outputs for inputs, often require expensive manual effort to label test data, possibly involving multiple experts to ensure labeling correctness. In this paper, we propose DeepGD, a black-box multi-objective test selection approach for DNN models. It reduces the cost of labeling by prioritizing the selection of test inputs with high fault revealing power from large unlabeled datasets. DeepGD not only selects test inputs with high uncertainty scores to trigger as many mispredicted inputs as possible but also maximizes the probability of revealing distinct faults in the DNN model by selecting diverse mispredicted inputs. The experimental results conducted on four widely used datasets and five DNN models show that in terms of fault-revealing ability: (1) White-box, coverage-based approaches fare poorly, (2) DeepGD outperforms existing black-box test selection approaches in terms of fault detection, and (3) DeepGD also leads to better guidance for DNN model retraining when using selected inputs to augment the training set.
♻ ☆ Extensible Machine Learning for Encrypted Network Traffic Application Labeling via Uncertainty Quantification
With the increasing prevalence of encrypted network traffic, cyber security analysts have been turning to machine learning (ML) techniques to elucidate the traffic on their networks. However, ML models can become stale as new traffic emerges that is outside of the distribution of the training set. In order to reliably adapt in this dynamic environment, ML models must additionally provide contextualized uncertainty quantification to their predictions, which has received little attention in the cyber security domain. Uncertainty quantification is necessary both to signal when the model is uncertain about which class to choose in its label assignment and when the traffic is not likely to belong to any pre-trained classes. We present a new, public dataset of network traffic that includes labeled, Virtual Private Network (VPN)-encrypted network traffic generated by 10 applications and corresponding to 5 application categories. We also present an ML framework that is designed to rapidly train with modest data requirements and provide both calibrated, predictive probabilities as well as an interpretable "out-of-distribution" (OOD) score to flag novel traffic samples. We describe calibrating OOD scores using p-values of the relative Mahalanobis distance. We demonstrate that our framework achieves an F1 score of 0.98 on our dataset and that it can extend to an enterprise network by testing the model: (1) on data from similar applications, (2) on dissimilar application traffic from an existing category, and (3) on application traffic from a new category. The model correctly flags uncertain traffic and, upon retraining, accurately incorporates the new data.
comment: Paper is 15 pages and has 10 figures. Published in IEEE Transactions on Artificial Intelligence (https://doi.org/10.1109/TAI.2023.3244168). For associated dataset, see https://www.ll.mit.edu/r-d/datasets/vpnnonvpn-network-application-traffic-dataset-vnat
♻ ☆ From Tensor Network Quantum States to Tensorial Recurrent Neural Networks
We show that any matrix product state (MPS) can be exactly represented by a recurrent neural network (RNN) with a linear memory update. We generalize this RNN architecture to 2D lattices using a multilinear memory update. It supports perfect sampling and wave function evaluation in polynomial time, and can represent an area law of entanglement entropy. Numerical evidence shows that it can encode the wave function using a bond dimension lower by orders of magnitude when compared to MPS, with an accuracy that can be systematically improved by increasing the bond dimension.
comment: 14 pages, 10 figures
♻ ☆ Discovering Closed-Loop Failures of Vision-Based Controllers via Reachability Analysis
Machine learning driven image-based controllers allow robotic systems to take intelligent actions based on the visual feedback from their environment. Understanding when these controllers might lead to system safety violations is important for their integration in safety-critical applications and engineering corrective safety measures for the system. Existing methods leverage simulation-based testing (or falsification) to find the failures of vision-based controllers, i.e., the visual inputs that lead to closed-loop safety violations. However, these techniques do not scale well to the scenarios involving high-dimensional and complex visual inputs, such as RGB images. In this work, we cast the problem of finding closed-loop vision failures as a Hamilton-Jacobi (HJ) reachability problem. Our approach blends simulation-based analysis with HJ reachability methods to compute an approximation of the backward reachable tube (BRT) of the system, i.e., the set of unsafe states for the system under vision-based controllers. Utilizing the BRT, we can tractably and systematically find the system states and corresponding visual inputs that lead to closed-loop failures. These visual inputs can be subsequently analyzed to find the input characteristics that might have caused the failure. Besides its scalability to high-dimensional visual inputs, an explicit computation of BRT allows the proposed approach to capture non-trivial system failures that are difficult to expose via random simulations. We demonstrate our framework on two case studies involving an RGB image-based neural network controller for (a) autonomous indoor navigation, and (b) autonomous aircraft taxiing.
♻ ☆ Probing Predictions on OOD Images via Nearest Categories
We study out-of-distribution (OOD) prediction behavior of neural networks when they classify images from unseen classes or corrupted images. To probe the OOD behavior, we introduce a new measure, nearest category generalization (NCG), where we compute the fraction of OOD inputs that are classified with the same label as their nearest neighbor in the training set. Our motivation stems from understanding the prediction patterns of adversarially robust networks, since previous work has identified unexpected consequences of training to be robust to norm-bounded perturbations. We find that robust networks have consistently higher NCG accuracy than natural training, even when the OOD data is much farther away than the robustness radius. This implies that the local regularization of robust training has a significant impact on the network's decision regions. We replicate our findings using many datasets, comparing new and existing training methods. Overall, adversarially robust networks resemble a nearest neighbor classifier when it comes to OOD data. Code available at https://github.com/yangarbiter/nearest-category-generalization.
comment: Accepted by Transactions on Machine Learning Research
♻ ☆ PASHA: Efficient HPO and NAS with Progressive Resource Allocation ICLR 2023
Hyperparameter optimization (HPO) and neural architecture search (NAS) are methods of choice to obtain the best-in-class machine learning models, but in practice they can be costly to run. When models are trained on large datasets, tuning them with HPO or NAS rapidly becomes prohibitively expensive for practitioners, even when efficient multi-fidelity methods are employed. We propose an approach to tackle the challenge of tuning machine learning models trained on large datasets with limited computational resources. Our approach, named PASHA, extends ASHA and is able to dynamically allocate maximum resources for the tuning procedure depending on the need. The experimental comparison shows that PASHA identifies well-performing hyperparameter configurations and architectures while consuming significantly fewer computational resources than ASHA.
comment: Accepted at ICLR 2023
♻ ☆ The autoregressive neural network architecture of the Boltzmann distribution of pairwise interacting spins systems
Generative Autoregressive Neural Networks (ARNN) have recently demonstrated exceptional results in image and language generation tasks, contributing to the growing popularity of generative models in both scientific and commercial applications. This work presents a physical interpretation of the ARNNs by reformulating the Boltzmann distribution of binary pairwise interacting systems into autoregressive form. The resulting ARNN architecture has weights and biases of its first layer corresponding to the Hamiltonian's couplings and external fields, featuring widely used structures like the residual connections and a recurrent architecture with clear physical meanings. However, the exponential growth, with system size, of the number of parameters of the hidden layers makes its direct application unfeasible. Nevertheless, its architecture's explicit formulation allows using statistical physics techniques to derive new ARNNs for specific systems. As examples, new effective ARNN architectures are derived from two well-known mean-field systems, the Curie-Weiss and Sherrington-Kirkpatrick models, showing superior performances in approximating the Boltzmann distributions of the corresponding physics model compared to other commonly used ARNN architectures. The connection established between the physics of the system and the ARNN architecture provides a way to derive new neural network architectures for different interacting systems and interpret existing ones from a physical perspective.
comment: 10 pages, 6 figure plus the Supplementary Information
♻ ☆ Tensor Train for Global Optimization Problems in Robotics
The convergence of many numerical optimization techniques is highly dependent on the initial guess given to the solver. To address this issue, we propose a novel approach that utilizes tensor methods to initialize existing optimization solvers near global optima. Our method does not require access to a database of good solutions. We first transform the cost function, which depends on both task parameters and optimization variables, into a probability density function. The joint probability distribution of the task parameters and optimization variables is approximated using the Tensor Train model which enables efficient conditioning and sampling. Unlike existing methods, we treat the task parameters as random variables and for a given task we generate samples for decision variables from the conditional distribution to initialize the optimization solver. Our method can produce multiple solutions for a given task from different modes when they exist. We first evaluate the approach on benchmark functions for numerical optimization that are hard to solve using gradient-based optimization solvers with a naive initialization. The results show that the proposed method can generate samples close to global optima and from multiple modes. We then demonstrate the generality and relevance of our framework to robotics by applying it to inverse kinematics with obstacles and motion planning problems with a 7-DoF manipulator.
comment: 26 pages, 21 figures
♻ ☆ Learning Hyper Label Model for Programmatic Weak Supervision ICLR 2023
To reduce the human annotation efforts, the programmatic weak supervision (PWS) paradigm abstracts weak supervision sources as labeling functions (LFs) and involves a label model to aggregate the output of multiple LFs to produce training labels. Most existing label models require a parameter learning step for each dataset. In this work, we present a hyper label model that (once learned) infers the ground-truth labels for each dataset in a single forward pass without dataset-specific parameter learning. The hyper label model approximates an optimal analytical (yet computationally intractable) solution of the ground-truth labels. We train the model on synthetic data generated in the way that ensures the model approximates the analytical optimal solution, and build the model upon Graph Neural Network (GNN) to ensure the model prediction being invariant (or equivariant) to the permutation of LFs (or data points). On 14 real-world datasets, our hyper label model outperforms the best existing methods in both accuracy (by 1.4 points on average) and efficiency (by six times on average). Our code is available at https://github.com/wurenzhi/hyper_label_model
comment: Published as a conference paper at ICLR 2023
♻ ☆ Stochastic Variable Metric Proximal Gradient with variance reduction for non-convex composite optimization
This paper introduces a novel algorithm, the Perturbed Proximal Preconditioned SPIDER algorithm (3P-SPIDER), designed to solve finite sum non-convex composite optimization. It is a stochastic Variable Metric Forward-Backward algorithm, which allows approximate preconditioned forward operator and uses a variable metric proximity operator as the backward operator; it also proposes a mini-batch strategy with variance reduction to address the finite sum setting. We show that 3P-SPIDER extends some Stochastic preconditioned Gradient Descent-based algorithms and some Incremental Expectation Maximization algorithms to composite optimization and to the case the forward operator can not be computed in closed form. We also provide an explicit control of convergence in expectation of 3P-SPIDER, and study its complexity in order to satisfy the epsilon-approximate stationary condition. Our results are the first to combine the composite non-convex optimization setting, a variance reduction technique to tackle the finite sum setting by using a minibatch strategy and, to allow deterministic or random approximations of the preconditioned forward operator. Finally, through an application to inference in a logistic regression model with random effects, we numerically compare 3P-SPIDER to other stochastic forward-backward algorithms and discuss the role of some design parameters of 3P-SPIDER.
comment: Statistics and Computing, In press
♻ ☆ SHIFT15M: Fashion-specific dataset for set-to-set matching with several distribution shifts
This paper addresses the problem of set-to-set matching, which involves matching two different sets of items based on some criteria, especially in the case of high-dimensional items like images. Although neural networks have been applied to solve this problem, most machine learning-based approaches assume that the training and test data follow the same distribution, which is not always true in real-world scenarios. To address this limitation, we introduce SHIFT15M, a dataset that can be used to evaluate set-to-set matching models when the distribution of data changes between training and testing. We conduct benchmark experiments that demonstrate the performance drop of naive methods due to distribution shift. Additionally, we provide software to handle the SHIFT15M dataset in a simple manner, with the URL for the software to be made available after publication of this manuscript. We believe proposed SHIFT15M dataset provide a valuable resource for evaluating set-to-set matching models under the distribution shift.
♻ ☆ Covid19 Reproduction Number: Credibility Intervals by Blockwise Proximal Monte Carlo Samplers
Monitoring the Covid19 pandemic constitutes a critical societal stake that received considerable research efforts. The intensity of the pandemic on a given territory is efficiently measured by the reproduction number, quantifying the rate of growth of daily new infections. Recently, estimates for the time evolution of the reproduction number were produced using an inverse problem formulation with a nonsmooth functional minimization. While it was designed to be robust to the limited quality of the Covid19 data (outliers, missing counts), the procedure lacks the ability to output credibility interval based estimates. This remains a severe limitation for practical use in actual pandemic monitoring by epidemiologists that the present work aims to overcome by use of Monte Carlo sampling. After interpretation of the nonsmooth functional into a Bayesian framework, several sampling schemes are tailored to adjust the nonsmooth nature of the resulting posterior distribution. The originality of the devised algorithms stems from combining a Langevin Monte Carlo sampling scheme with Proximal operators. Performance of the new algorithms in producing relevant credibility intervals for the reproduction number estimates and denoised counts are compared. Assessment is conducted on real daily new infection counts made available by the Johns Hopkins University. The interest of the devised monitoring tools are illustrated on Covid19 data from several different countries.
♻ ☆ CaDM: Codec-aware Diffusion Modeling for Neural-enhanced Video Streaming
Recent years have witnessed the dramatic growth of Internet video traffic, where the video bitstreams are often compressed and delivered in low quality to fit the streamer's uplink bandwidth. To alleviate the quality degradation, it comes the rise of Neural-enhanced Video Streaming (NVS), which shows great prospects for recovering low-quality videos by mostly deploying neural super-resolution (SR) on the media server. Despite its benefit, we reveal that current mainstream works with SR enhancement have not achieved the desired rate-distortion trade-off between bitrate saving and quality restoration, due to: (1) overemphasizing the enhancement on the decoder side while omitting the co-design of encoder, (2) limited generative capacity to recover high-fidelity perceptual details, and (3) optimizing the compression-and-restoration pipeline from the resolution perspective solely, without considering color bit-depth. Aiming at overcoming these limitations, we are the first to conduct an encoder-decoder (i.e., codec) synergy by leveraging the inherent visual-generative property of diffusion models. Specifically, we present the Codec-aware Diffusion Modeling (CaDM), a novel NVS paradigm to significantly reduce streaming delivery bitrates while holding pretty higher restoration capacity over existing methods. First, CaDM improves the encoder's compression efficiency by simultaneously reducing resolution and color bit-depth of video frames. Second, CaDM empowers the decoder with high-quality enhancement by making the denoising diffusion restoration aware of encoder's resolution-color conditions. Evaluation on public cloud services with OpenMMLab benchmarks shows that CaDM effectively saves up to 5.12 - 21.44 times bitrates based on common video standards and achieves much better recovery quality (e.g., FID of 0.61) over state-of-the-art neural-enhancing methods.
♻ ☆ Comparison of semi-supervised deep learning algorithms for audio classification
In this article, we adapted five recent SSL methods to the task of audio classification. The first two methods, namely Deep Co-Training (DCT) and Mean Teacher (MT), involve two collaborative neural networks. The three other algorithms, called MixMatch (MM), ReMixMatch (RMM), and FixMatch (FM), are single-model methods that rely primarily on data augmentation strategies. Using the Wide-ResNet-28-2 architecture in all our experiments, 10% of labeled data and the remaining 90% as unlabeled data for training, we first compare the error rates of the five methods on three standard benchmark audio datasets: Environmental Sound Classification (ESC-10), UrbanSound8K (UBS8K), and Google Speech Commands (GSC). In all but one cases, MM, RMM, and FM outperformed MT and DCT significantly, MM and RMM being the best methods in most experiments. On UBS8K and GSC, MM achieved 18.02% and 3.25% error rate (ER), respectively, outperforming models trained with 100% of the available labeled data, which reached 23.29% and 4.94%, respectively. RMM achieved the best results on ESC-10 (12.00% ER), followed by FM which reached 13.33%. Second, we explored adding the mixup augmentation, used in MM and RMM, to DCT, MT, and FM. In almost all cases, mixup brought consistent gains. For instance, on GSC, FM reached 4.44% and 3.31% ER without and with mixup. Our PyTorch code will be made available upon paper acceptance at https:// github. com/ Labbe ti/ SSLH.
comment: 9 pages, 5 figures, 5 tables. This is the version 3 of the paper. Contains minor fixes compared to the EURASIP one (which is the version 2 of the paper)
♻ ☆ GLCC: A General Framework for Graph-Level Clustering AAAI
This paper studies the problem of graph-level clustering, which is a novel yet challenging task. This problem is critical in a variety of real-world applications such as protein clustering and genome analysis in bioinformatics. Recent years have witnessed the success of deep clustering coupled with graph neural networks (GNNs). However, existing methods focus on clustering among nodes given a single graph, while exploring clustering on multiple graphs is still under-explored. In this paper, we propose a general graph-level clustering framework named Graph-Level Contrastive Clustering (GLCC) given multiple graphs. Specifically, GLCC first constructs an adaptive affinity graph to explore instance- and cluster-level contrastive learning (CL). Instance-level CL leverages graph Laplacian based contrastive loss to learn clustering-friendly representations while cluster-level CL captures discriminative cluster representations incorporating neighbor information of each sample. Moreover, we utilize neighbor-aware pseudo-labels to reward the optimization of representation learning. The two steps can be alternatively trained to collaborate and benefit each other. Experiments on a range of well-known datasets demonstrate the superiority of our proposed GLCC over competitive baselines.
comment: Accepted by Proceedings of the AAAI Conference on Artificial Intelligence (AAAI 2023)
♻ ☆ Variance Reduction is an Antidote to Byzantines: Better Rates, Weaker Assumptions and Communication Compression as a Cherry on the Top ICLR 2023
Byzantine-robustness has been gaining a lot of attention due to the growth of the interest in collaborative and federated learning. However, many fruitful directions, such as the usage of variance reduction for achieving robustness and communication compression for reducing communication costs, remain weakly explored in the field. This work addresses this gap and proposes Byz-VR-MARINA - a new Byzantine-tolerant method with variance reduction and compression. A key message of our paper is that variance reduction is key to fighting Byzantine workers more effectively. At the same time, communication compression is a bonus that makes the process more communication efficient. We derive theoretical convergence guarantees for Byz-VR-MARINA outperforming previous state-of-the-art for general non-convex and Polyak-Lojasiewicz loss functions. Unlike the concurrent Byzantine-robust methods with variance reduction and/or compression, our complexity results are tight and do not rely on restrictive assumptions such as boundedness of the gradients or limited compression. Moreover, we provide the first analysis of a Byzantine-tolerant method supporting non-uniform sampling of stochastic gradients. Numerical experiments corroborate our theoretical findings.
comment: ICLR 2023. 42 pages, 8 figures. Changes in v2: few typos and inaccuracies were fixed, more clarifications were added. Changes in v3: ICLR formatting was applied, additional experiments were added (Appendix B.4-B.5) and extra discussion of the results was added to Appendix E.5. Code: https://github.com/SamuelHorvath/VR_Byzantine
♻ ☆ Research on Efficient Fuzzy Clustering Method Based on Local Fuzzy Granular balls
In recent years, the problem of fuzzy clustering has been widely concerned. The membership iteration of existing methods is mostly considered globally, which has considerable problems in noisy environments, and iterative calculations for clusters with a large number of different sample sizes are not accurate and efficient. In this paper, starting from the strategy of large-scale priority, the data is fuzzy iterated using granular-balls, and the membership degree of data only considers the two granular-balls where it is located, thus improving the efficiency of iteration. The formed fuzzy granular-balls set can use more processing methods in the face of different data scenarios, which enhances the practicability of fuzzy clustering calculations.
♻ ☆ Stochastic Gradient Descent-Ascent: Unified Theory and New Efficient Methods AISTATS 2023
Stochastic Gradient Descent-Ascent (SGDA) is one of the most prominent algorithms for solving min-max optimization and variational inequalities problems (VIP) appearing in various machine learning tasks. The success of the method led to several advanced extensions of the classical SGDA, including variants with arbitrary sampling, variance reduction, coordinate randomization, and distributed variants with compression, which were extensively studied in the literature, especially during the last few years. In this paper, we propose a unified convergence analysis that covers a large variety of stochastic gradient descent-ascent methods, which so far have required different intuitions, have different applications and have been developed separately in various communities. A key to our unified framework is a parametric assumption on the stochastic estimates. Via our general theoretical framework, we either recover the sharpest known rates for the known special cases or tighten them. Moreover, to illustrate the flexibility of our approach we develop several new variants of SGDA such as a new variance-reduced method (L-SVRGDA), new distributed methods with compression (QSGDA, DIANA-SGDA, VR-DIANA-SGDA), and a new method with coordinate randomization (SEGA-SGDA). Although variants of the new methods are known for solving minimization problems, they were never considered or analyzed for solving min-max problems and VIPs. We also demonstrate the most important properties of the new methods through extensive numerical experiments.
comment: AISTATS 2023. 65 pages, 5 figures, 3 tables. Changes in v2: new results were added (Theorem 2.5 and its corollaries), few typos were fixed, more clarifications were added. Changes in v3: AISTATS formatting was applied, small clarifications were added. Code: https://github.com/hugobb/sgda
♻ ☆ Generation of non-stationary stochastic fields using Generative Adversarial Networks
In the context of generating geological facies conditioned on observed data, samples corresponding to all possible conditions are not generally available in the training set and hence the generation of these realizations depends primary on the generalization capability of the trained generative model. The problem becomes more complex when applied on non-stationary fields. In this work, we investigate the problem of using Generative Adversarial Networks (GANs) models to generate non-stationary geological channelized patterns and examine the models generalization capability at new spatial modes that were never seen in the given training set. The developed training method based on spatial-conditioning allowed for effective learning of the correlation between the spatial conditions (i.e. non-stationary maps) and the realizations implicitly without using additional loss terms or solving optimization problems for every new given data after training. In addition, our models can be trained on 2D and 3D samples. The results on real and artificial datasets show that we were able to generate geologically-plausible realizations beyond the training samples and with a strong correlation with the target maps.
♻ ☆ Causal Inference Despite Limited Global Confounding via Mixture Models
A Bayesian Network is a directed acyclic graph (DAG) on a set of $n$ random variables (the vertices); a Bayesian Network Distribution (BND) is a probability distribution on the random variables that is Markovian on the graph. A finite $k$-mixture of such models is graphically represented by a larger graph which has an additional "hidden" (or "latent") random variable $U$, ranging in $\{1,\ldots,k\}$, and a directed edge from $U$ to every other vertex. Models of this type are fundamental to causal inference, where $U$ models an unobserved confounding effect of multiple populations, obscuring the causal relationships in the observable DAG. By solving the mixture problem and recovering the joint probability distribution on $U$, traditionally unidentifiable causal relationships become identifiable. Using a reduction to the more well-studied "product" case on empty graphs, we give the first algorithm to learn mixtures of non-empty DAGs.
comment: Paper to appear in CLEAR 2023
♻ ☆ Domain Adaptation of Transformer-Based Models using Unlabeled Data for Relevance and Polarity Classification of German Customer Feedback
Understanding customer feedback is becoming a necessity for companies to identify problems and improve their products and services. Text classification and sentiment analysis can play a major role in analyzing this data by using a variety of machine and deep learning approaches. In this work, different transformer-based models are utilized to explore how efficient these models are when working with a German customer feedback dataset. In addition, these pre-trained models are further analyzed to determine if adapting them to a specific domain using unlabeled data can yield better results than off-the-shelf pre-trained models. To evaluate the models, two downstream tasks from the GermEval 2017 are considered. The experimental results show that transformer-based models can reach significant improvements compared to a fastText baseline and outperform the published scores and previous models. For the subtask Relevance Classification, the best models achieve a micro-averaged $F1$-Score of 96.1 % on the first test set and 95.9 % on the second one, and a score of 85.1 % and 85.3 % for the subtask Polarity Classification.
comment: Complete
♻ ☆ Learning Imbalanced Data with Vision Transformers CVPR 2023
The real-world data tends to be heavily imbalanced and severely skew the data-driven deep neural networks, which makes Long-Tailed Recognition (LTR) a massive challenging task. Existing LTR methods seldom train Vision Transformers (ViTs) with Long-Tailed (LT) data, while the off-the-shelf pretrain weight of ViTs always leads to unfair comparisons. In this paper, we systematically investigate the ViTs' performance in LTR and propose LiVT to train ViTs from scratch only with LT data. With the observation that ViTs suffer more severe LTR problems, we conduct Masked Generative Pretraining (MGP) to learn generalized features. With ample and solid evidence, we show that MGP is more robust than supervised manners. In addition, Binary Cross Entropy (BCE) loss, which shows conspicuous performance with ViTs, encounters predicaments in LTR. We further propose the balanced BCE to ameliorate it with strong theoretical groundings. Specially, we derive the unbiased extension of Sigmoid and compensate extra logit margins to deploy it. Our Bal-BCE contributes to the quick convergence of ViTs in just a few epochs. Extensive experiments demonstrate that with MGP and Bal-BCE, LiVT successfully trains ViTs well without any additional data and outperforms comparable state-of-the-art methods significantly, e.g., our ViT-B achieves 81.0% Top-1 accuracy in iNaturalist 2018 without bells and whistles. Code is available at https://github.com/XuZhengzhuo/LiVT.
comment: Accepted to CVPR 2023, camera-ready version; Code: https://github.com/XuZhengzhuo/LiVT
♻ ☆ Towards Inferential Reproducibility of Machine Learning Research ICLR 2023
Reliability of machine learning evaluation -- the consistency of observed evaluation scores across replicated model training runs -- is affected by several sources of nondeterminism which can be regarded as measurement noise. Current tendencies to remove noise in order to enforce reproducibility of research results neglect inherent nondeterminism at the implementation level and disregard crucial interaction effects between algorithmic noise factors and data properties. This limits the scope of conclusions that can be drawn from such experiments. Instead of removing noise, we propose to incorporate several sources of variance, including their interaction with data properties, into an analysis of significance and reliability of machine learning evaluation, with the aim to draw inferences beyond particular instances of trained models. We show how to use linear mixed effects models (LMEMs) to analyze performance evaluation scores, and to conduct statistical inference with a generalized likelihood ratio test (GLRT). This allows us to incorporate arbitrary sources of noise like meta-parameter variations into statistical significance testing, and to assess performance differences conditional on data properties. Furthermore, a variance component analysis (VCA) enables the analysis of the contribution of noise sources to overall variance and the computation of a reliability coefficient by the ratio of substantial to total variance.
comment: Published at ICLR 2023 (see https://openreview.net/pdf?id=li4GQCQWkv)
♻ ☆ Feature Selection for Forecasting
This work investigates the importance of feature selection for improving the forecasting performance of machine learning algorithms for financial data. Artificial neural networks (ANN), convolutional neural networks (CNN), long-short term memory (LSTM) networks, as well as linear models were applied for forecasting purposes. The Feature Selection with Annealing (FSA) algorithm was used to select the features from about 1000 possible predictors obtained from 26 technical indicators with specific periods and their lags. In addition to this, the Boruta feature selection algorithm was applied as a baseline feature selection method. The dependent variables consisted of daily logarithmic returns and daily trends of ten financial data sets, including cryptocurrency and different stocks. Experiments indicate that the FSA algorithm increased the performance of ML models regardless of the problem type. The FSA hybrid machine learning models showed better performance in 10 out of 10 data sets for regression and 8 out of 10 data sets for classification. None of the hybrid Boruta models outperformed the hybrid FSA models. However, the BORCNN model performance was comparable to the best model for 4 out of 10 data sets for regression estimates. BOR-LR and BOR-CNN models showed comparable performance with the best hybrid FSA models in 2 out of 10 datasets for classification. FSA was observed to improve the model performance in both better performance metrics as well as a decreased computation time by providing a lower dimensional input feature space.
comment: 21 pages, 2 figures and 12 tables
♻ ☆ SpReME: Sparse Regression for Multi-Environment Dynamic Systems
Learning dynamical systems is a promising avenue for scientific discoveries. However, capturing the governing dynamics in multiple environments still remains a challenge: model-based approaches rely on the fidelity of assumptions made for a single environment, whereas data-driven approaches based on neural networks are often fragile on extrapolating into the future. In this work, we develop a method of sparse regression dubbed SpReME to discover the major dynamics that underlie multiple environments. Specifically, SpReME shares a sparse structure of ordinary differential equation (ODE) across different environments in common while allowing each environment to keep the coefficients of ODE terms independently. We demonstrate that the proposed model captures the correct dynamics from multiple environments over four different dynamic systems with improved prediction performance.
comment: The code is available at https://github.com/ml-postech/SpReME
♻ ☆ Extending DNN-based Multiplicative Masking to Deep Subband Filtering for Improved Dereverberation
In this paper, we present a scheme for extending deep neural network-based multiplicative maskers to deep subband filters for speech restoration in the time-frequency domain. The resulting method can be generically applied to any deep neural network providing masks in the time-frequency domain, while requiring only few more trainable parameters and a computational overhead that is negligible for state-of-the-art neural networks. We demonstrate that the resulting deep subband filtering scheme outperforms multiplicative masking for dereverberation, while leaving the denoising performance virtually the same. We argue that this is because deep subband filtering in the time-frequency domain fits the subband approximation often assumed in the dereverberation literature, whereas multiplicative masking corresponds to the narrowband approximation generally employed in denoising.
♻ ☆ Leveraging Heteroscedastic Uncertainty in Learning Complex Spectral Mapping for Single-channel Speech Enhancement ICASSP 2023
Most speech enhancement (SE) models learn a point estimate and do not make use of uncertainty estimation in the learning process. In this paper, we show that modeling heteroscedastic uncertainty by minimizing a multivariate Gaussian negative log-likelihood (NLL) improves SE performance at no extra cost. During training, our approach augments a model learning complex spectral mapping with a temporary submodel to predict the covariance of the enhancement error at each time-frequency bin. Due to unrestricted heteroscedastic uncertainty, the covariance introduces an undersampling effect, detrimental to SE performance. To mitigate undersampling, our approach inflates the uncertainty lower bound and weights each loss component with their uncertainty, effectively compensating severely undersampled components with more penalties. Our multivariate setting reveals common covariance assumptions such as scalar and diagonal matrices. By weakening these assumptions, we show that the NLL achieves superior performance compared to popular loss functions including the mean squared error (MSE), mean absolute error (MAE), and scale-invariant signal-to-distortion ratio (SI-SDR).
comment: 5 pages. Accepted at ICASSP 2023
♻ ☆ Grounding Language with Visual Affordances over Unstructured Data ICRA
Recent works have shown that Large Language Models (LLMs) can be applied to ground natural language to a wide variety of robot skills. However, in practice, learning multi-task, language-conditioned robotic skills typically requires large-scale data collection and frequent human intervention to reset the environment or help correcting the current policies. In this work, we propose a novel approach to efficiently learn general-purpose language-conditioned robot skills from unstructured, offline and reset-free data in the real world by exploiting a self-supervised visuo-lingual affordance model, which requires annotating as little as 1% of the total data with language. We evaluate our method in extensive experiments both in simulated and real-world robotic tasks, achieving state-of-the-art performance on the challenging CALVIN benchmark and learning over 25 distinct visuomotor manipulation tasks with a single policy in the real world. We find that when paired with LLMs to break down abstract natural language instructions into subgoals via few-shot prompting, our method is capable of completing long-horizon, multi-tier tasks in the real world, while requiring an order of magnitude less data than previous approaches. Code and videos are available at http://hulc2.cs.uni-freiburg.de
comment: Accepted at the 2023 IEEE International Conference on Robotics and Automation (ICRA). Project website: http://hulc2.cs.uni-freiburg.de
♻ ☆ Visual Language Maps for Robot Navigation ICRA
Grounding language to the visual observations of a navigating agent can be performed using off-the-shelf visual-language models pretrained on Internet-scale data (e.g., image captions). While this is useful for matching images to natural language descriptions of object goals, it remains disjoint from the process of mapping the environment, so that it lacks the spatial precision of classic geometric maps. To address this problem, we propose VLMaps, a spatial map representation that directly fuses pretrained visual-language features with a 3D reconstruction of the physical world. VLMaps can be autonomously built from video feed on robots using standard exploration approaches and enables natural language indexing of the map without additional labeled data. Specifically, when combined with large language models (LLMs), VLMaps can be used to (i) translate natural language commands into a sequence of open-vocabulary navigation goals (which, beyond prior work, can be spatial by construction, e.g., "in between the sofa and TV" or "three meters to the right of the chair") directly localized in the map, and (ii) can be shared among multiple robots with different embodiments to generate new obstacle maps on-the-fly (by using a list of obstacle categories). Extensive experiments carried out in simulated and real world environments show that VLMaps enable navigation according to more complex language instructions than existing methods. Videos are available at https://vlmaps.github.io.
comment: Accepted at the 2023 IEEE International Conference on Robotics and Automation (ICRA). Project page: https://vlmaps.github.io
♻ ☆ Goal-Conditioned Q-Learning as Knowledge Distillation AAAI 2023
Many applications of reinforcement learning can be formalized as goal-conditioned environments, where, in each episode, there is a "goal" that affects the rewards obtained during that episode but does not affect the dynamics. Various techniques have been proposed to improve performance in goal-conditioned environments, such as automatic curriculum generation and goal relabeling. In this work, we explore a connection between off-policy reinforcement learning in goal-conditioned settings and knowledge distillation. In particular: the current Q-value function and the target Q-value estimate are both functions of the goal, and we would like to train the Q-value function to match its target for all goals. We therefore apply Gradient-Based Attention Transfer (Zagoruyko and Komodakis 2017), a knowledge distillation technique, to the Q-function update. We empirically show that this can improve the performance of goal-conditioned off-policy reinforcement learning when the space of goals is high-dimensional. We also show that this technique can be adapted to allow for efficient learning in the case of multiple simultaneous sparse goals, where the agent can attain a reward by achieving any one of a large set of objectives, all specified at test time. Finally, to provide theoretical support, we give examples of classes of environments where (under some assumptions) standard off-policy algorithms such as DDPG require at least O(d^2) replay buffer transitions to learn an optimal policy, while our proposed technique requires only O(d) transitions, where d is the dimensionality of the goal and state space. Code is available at https://github.com/alevine0/ReenGAGE.
comment: AAAI 2023 Accepted paper
♻ ☆ Mean-Semivariance Policy Optimization via Risk-Averse Reinforcement Learning
Keeping risk under control is often more crucial than maximizing expected rewards in real-world decision-making situations, such as finance, robotics, autonomous driving, etc. The most natural choice of risk measures is variance, which penalizes the upside volatility as much as the downside part. Instead, the (downside) semivariance, which captures the negative deviation of a random variable under its mean, is more suitable for risk-averse proposes. This paper aims at optimizing the mean-semivariance (MSV) criterion in reinforcement learning w.r.t. steady reward distribution. Since semivariance is time-inconsistent and does not satisfy the standard Bellman equation, the traditional dynamic programming methods are inapplicable to MSV problems directly. To tackle this challenge, we resort to Perturbation Analysis (PA) theory and establish the performance difference formula for MSV. We reveal that the MSV problem can be solved by iteratively solving a sequence of RL problems with a policy-dependent reward function. Further, we propose two on-policy algorithms based on the policy gradient theory and the trust region method. Finally, we conduct diverse experiments from simple bandit problems to continuous control tasks in MuJoCo, which demonstrate the effectiveness of our proposed methods.
comment: Accecpted by Journal of Artificial Intelligence Research
♻ ☆ ChatGPT: Beginning of an End of Manual Linguistic Data Annotation? Use Case of Automatic Genre Identification
ChatGPT has shown strong capabilities in natural language generation tasks, which naturally leads researchers to explore where its abilities end. In this paper, we examine whether ChatGPT can be used for zero-shot text classification, more specifically, automatic genre identification. We compare ChatGPT with a multilingual XLM-RoBERTa language model that was fine-tuned on datasets, manually annotated with genres. The models are compared on test sets in two languages: English and Slovenian. Results show that ChatGPT outperforms the fine-tuned model when applied to the dataset which was not seen before by either of the models. Even when applied on Slovenian language as an under-resourced language, ChatGPT's performance is no worse than when applied to English. However, if the model is fully prompted in Slovenian, the performance drops significantly, showing the current limitations of ChatGPT usage on smaller languages. The presented results lead us to questioning whether this is the beginning of an end of laborious manual annotation campaigns even for smaller languages, such as Slovenian.
♻ ☆ Configurable calorimeter simulation for AI applications
A configurable calorimeter simulation for AI (COCOA) applications is presented, based on the Geant4 toolkit and interfaced with the Pythia event generator. This open-source project is aimed to support the development of machine learning algorithms in high energy physics that rely on realistic particle shower descriptions, such as reconstruction, fast simulation, and low-level analysis. Specifications such as the granularity and material of its nearly hermetic geometry are user-configurable. The tool is supplemented with simple event processing including topological clustering, jet algorithms, and a nearest-neighbors graph construction. Formatting is also provided to visualise events using the Phoenix event display software.
comment: 9 pages, 11 figures
♻ ☆ Dimension-reduced KRnet maps for high-dimensional Bayesian inverse problems
We present a dimension-reduced KRnet map approach (DR-KRnet) for high-dimensional Bayesian inverse problems, which is based on an explicit construction of a map that pushes forward the prior measure to the posterior measure in the latent space. Our approach consists of two main components: data-driven VAE prior and density approximation of the posterior of the latent variable. In reality, it may not be trivial to initialize a prior distribution that is consistent with available prior data; in other words, the complex prior information is often beyond simple hand-crafted priors. We employ variational autoencoder (VAE) to approximate the underlying distribution of the prior dataset, which is achieved through a latent variable and a decoder. Using the decoder provided by the VAE prior, we reformulate the problem in a low-dimensional latent space. In particular, we seek an invertible transport map given by KRnet to approximate the posterior distribution of the latent variable. Moreover, an efficient physics-constrained surrogate model without any labeled data is constructed to reduce the computational cost of solving both forward and adjoint problems involved in likelihood computation. With numerical experiments, we demonstrate the accuracy and efficiency of DR-KRnet for high-dimensional Bayesian inverse problems.
♻ ☆ Improving Covariance-Regularized Discriminant Analysis for EHR-based Predictive Analytics of Diseases
Linear Discriminant Analysis (LDA) is a well-known technique for feature extraction and dimension reduction. The performance of classical LDA, however, significantly degrades on the High Dimension Low Sample Size (HDLSS) data for the ill-posed inverse problem. Existing approaches for HDLSS data classification typically assume the data in question are with Gaussian distribution and deal the HDLSS classification problem with regularization. However, these assumptions are too strict to hold in many emerging real-life applications, such as enabling personalized predictive analysis using Electronic Health Records (EHRs) data collected from an extremely limited number of patients who have been diagnosed with or without the target disease for prediction. In this paper, we revised the problem of predictive analysis of disease using personal EHR data and LDA classifier. To fill the gap, in this paper, we first studied an analytical model that understands the accuracy of LDA for classifying data with arbitrary distribution. The model gives a theoretical upper bound of LDA error rate that is controlled by two factors: (1) the statistical convergence rate of (inverse) covariance matrix estimators and (2) the divergence of the training/testing datasets to fitted distributions. To this end, we could lower the error rate by balancing the two factors for better classification performance. Hereby, we further proposed a novel LDA classifier De-Sparse that leverages De-sparsified Graphical Lasso to improve the estimation of LDA, which outperforms state-of-the-art LDA approaches developed for HDLSS data. Such advances and effectiveness are further demonstrated by both theoretical analysis and extensive experiments on EHR datasets.
comment: Sijia Yang wrote the manuscript into to the current version
♻ ☆ RLx2: Training a Sparse Deep Reinforcement Learning Model from Scratch ICLR 2023
Training deep reinforcement learning (DRL) models usually requires high computation costs. Therefore, compressing DRL models possesses immense potential for training acceleration and model deployment. However, existing methods that generate small models mainly adopt the knowledge distillation-based approach by iteratively training a dense network. As a result, the training process still demands massive computing resources. Indeed, sparse training from scratch in DRL has not been well explored and is particularly challenging due to non-stationarity in bootstrap training. In this work, we propose a novel sparse DRL training framework, "the Rigged Reinforcement Learning Lottery" (RLx2), which builds upon gradient-based topology evolution and is capable of training a sparse DRL model based entirely on a sparse network. Specifically, RLx2 introduces a novel multi-step TD target mechanism with a dynamic-capacity replay buffer to achieve robust value learning and efficient topology exploration in sparse models. It also reaches state-of-the-art sparse training performance in several tasks, showing 7.5\times-20\times model compression with less than 3% performance degradation and up to 20\times and 50\times FLOPs reduction for training and inference, respectively.
comment: ICLR 2023 spotlight
♻ ☆ PixCUE: Joint Uncertainty Estimation and Image Reconstruction in MRI using Deep Pixel Classification
Deep learning (DL) models are capable of successfully exploiting latent representations in MR data and have become state-of-the-art for accelerated MRI reconstruction. However, undersampling the measurements in k-space as well as the over- or under-parameterized and non-transparent nature of DL make these models exposed to uncertainty. Consequently, uncertainty estimation has become a major issue in DL MRI reconstruction. To estimate uncertainty, Monte Carlo (MC) inference techniques have become a common practice where multiple reconstructions are utilized to compute the variance in reconstruction as a measurement of uncertainty. However, these methods demand high computational costs as they require multiple inferences through the DL model. To this end, we introduce a method to estimate uncertainty during MRI reconstruction using a pixel classification framework. The proposed method, PixCUE (stands for Pixel Classification Uncertainty Estimation) produces the reconstructed image along with an uncertainty map during a single forward pass through the DL model. We demonstrate that this approach generates uncertainty maps that highly correlate with the reconstruction errors with respect to various MR imaging sequences and under numerous adversarial conditions. We also show that the estimated uncertainties are correlated to that of the conventional MC method. We further provide an empirical relationship between the uncertainty estimations using PixCUE and well-established reconstruction metrics such as NMSE, PSNR, and SSIM. We conclude that PixCUE is capable of reliably estimating the uncertainty in MRI reconstruction with a minimum additional computational cost.
comment: 19 pages, 7 figures, 1 table
♻ ☆ CleanCLIP: Mitigating Data Poisoning Attacks in Multimodal Contrastive Learning
Multimodal contrastive pretraining has been used to train multimodal representation models, such as CLIP, on large amounts of paired image-text data. However, previous studies have revealed that such models are vulnerable to backdoor attacks. Specifically, when trained on backdoored examples, CLIP learns spurious correlations between the embedded backdoor trigger and the target label, aligning their representations in the joint embedding space. Injecting even a small number of poisoned examples, such as 75 examples in 3 million pretraining data, can significantly manipulate the model's behavior, making it difficult to detect or unlearn such correlations. To address this issue, we propose CleanCLIP, a finetuning framework that weakens the learned spurious associations introduced by backdoor attacks by independently re-aligning the representations for individual modalities. We demonstrate that unsupervised finetuning using a combination of multimodal contrastive and unimodal self-supervised objectives for individual modalities can significantly reduce the impact of the backdoor attack. Additionally, we show that supervised finetuning on task-specific labeled image data removes the backdoor trigger from the CLIP vision encoder. We show empirically that CleanCLIP maintains model performance on benign examples while erasing a range of backdoor attacks on multimodal contrastive learning.
comment: 20 pages, 7 figures, 8 tables
♻ ☆ Graph Positional Encoding via Random Feature Propagation
Two main families of node feature augmentation schemes have been explored for enhancing GNNs: random features and spectral positional encoding. Surprisingly, however, there is still no clear understanding of the relation between these two augmentation schemes. Here we propose a novel family of positional encoding schemes which draws a link between the above two approaches and improves over both. The new approach, named Random Feature Propagation (RFP), is inspired by the power iteration method and its generalizations. It concatenates several intermediate steps of an iterative algorithm for computing the dominant eigenvectors of a propagation matrix, starting from random node features. Notably, these propagation steps are based on graph-dependent propagation operators that can be either predefined or learned. We explore the theoretical and empirical benefits of RFP. First, we provide theoretical justifications for using random features, for incorporating early propagation steps, and for using multiple random initializations. Then, we empirically demonstrate that RFP significantly outperforms both spectral PE and random features in multiple node classification and graph classification benchmarks.
♻ ☆ Switchable Representation Learning Framework with Self-compatibility CVPR 2023
Real-world visual search systems involve deployments on multiple platforms with different computing and storage resources. Deploying a unified model that suits the minimal-constrain platforms leads to limited accuracy. It is expected to deploy models with different capacities adapting to the resource constraints, which requires features extracted by these models to be aligned in the metric space. The method to achieve feature alignments is called ``compatible learning''. Existing research mainly focuses on the one-to-one compatible paradigm, which is limited in learning compatibility among multiple models. We propose a Switchable representation learning Framework with Self-Compatibility (SFSC). SFSC generates a series of compatible sub-models with different capacities through one training process. The optimization of sub-models faces gradients conflict, and we mitigate this problem from the perspective of the magnitude and direction. We adjust the priorities of sub-models dynamically through uncertainty estimation to co-optimize sub-models properly. Besides, the gradients with conflicting directions are projected to avoid mutual interference. SFSC achieves state-of-the-art performance on the evaluated datasets.
comment: Accepted by CVPR 2023
♻ ☆ Drugs Resistance Analysis from Scarce Health Records via Multi-task Graph Representation
Clinicians prescribe antibiotics by looking at the patient's health record with an experienced eye. However, the therapy might be rendered futile if the patient has drug resistance. Determining drug resistance requires time-consuming laboratory-level testing while applying clinicians' heuristics in an automated way is difficult due to the categorical or binary medical events that constitute health records. In this paper, we propose a novel framework for rapid clinical intervention by viewing health records as graphs whose nodes are mapped from medical events and edges as correspondence between events in given a time window. A novel graph-based model is then proposed to extract informative features and yield automated drug resistance analysis from those high-dimensional and scarce graphs. The proposed method integrates multi-task learning into a common feature extracting graph encoder for simultaneous analyses of multiple drugs as well as stabilizing learning. On a massive dataset comprising over 110,000 patients with urinary tract infections, we verify the proposed method is capable of attaining superior performance on the drug resistance prediction problem. Furthermore, automated drug recommendations resemblant to laboratory-level testing can also be made based on the model resistance analysis.
comment: 12 pages, 5 figures
♻ ☆ AutoFR: Automated Filter Rule Generation for Adblocking
Adblocking relies on filter lists, which are manually curated and maintained by a community of filter list authors. Filter list curation is a laborious process that does not scale well to a large number of sites or over time. In this paper, we introduce AutoFR, a reinforcement learning framework to fully automate the process of filter rule creation and evaluation for sites of interest. We design an algorithm based on multi-arm bandits to generate filter rules that block ads while controlling the trade-off between blocking ads and avoiding visual breakage. We test AutoFR on thousands of sites and we show that it is efficient: it takes only a few minutes to generate filter rules for a site of interest. AutoFR is effective: it generates filter rules that can block 86% of the ads, as compared to 87% by EasyList, while achieving comparable visual breakage. Furthermore, AutoFR generates filter rules that generalize well to new sites. We envision that AutoFR can assist the adblocking community in filter rule generation at scale.
comment: 16 pages with 13 figures, 3 tables, 1 algorithm. 3.5 pages of references. Appendices include 10 pages of appendices with 11 figures and 3 tables
♻ ☆ RETEXO: Scalable Neural Network Training over Distributed Graphs
Graph neural networks offer a promising approach to supervised learning over graph data. Graph data, especially when it is privacy-sensitive or too large to train on centrally, is often stored partitioned across disparate processing units (clients) which want to minimize the communication costs during collaborative training. The fully-distributed setup takes such partitioning to its extreme, wherein features of only a single node and its adjacent edges are kept locally with one client processor. Existing GNNs are not architected for training in such setups and incur prohibitive costs therein. We propose RETEXO, a novel transformation of existing GNNs that improves the communication efficiency during training in the fully-distributed setup. We experimentally confirm that RETEXO offers up to 6 orders of magnitude better communication efficiency even when training shallow GNNs, with a minimal trade-off in accuracy for supervised node classification tasks.
♻ ☆ Can Membership Inferencing be Refuted?
Membership inference (MI) attack is currently the most popular test for measuring privacy leakage in machine learning models. Given a machine learning model, a data point and some auxiliary information, the goal of an MI attack is to determine whether the data point was used to train the model. In this work, we study the reliability of membership inference attacks in practice. Specifically, we show that a model owner can plausibly refute the result of a membership inference test on a data point $x$ by constructing a proof of repudiation that proves that the model was trained without $x$. We design efficient algorithms to construct proofs of repudiation for all data points of the training dataset. Our empirical evaluation demonstrates the practical feasibility of our algorithm by constructing proofs of repudiation for popular machine learning models on MNIST and CIFAR-10. Consequently, our results call for a re-evaluation of the implications of membership inference attacks in practice.
♻ ☆ Line Graph Contrastive Learning for Link Prediction
Link prediction tasks focus on predicting possible future connections. Most existing researches measure the likelihood of links by different similarity scores on node pairs and predict links between nodes. However, the similarity-based approaches have some challenges in information loss on nodes and generalization ability on similarity indexes. To address the above issues, we propose a Line Graph Contrastive Learning(LGCL) method to obtain rich information with multiple perspectives. LGCL obtains a subgraph view by h-hop subgraph sampling with target node pairs. After transforming the sampled subgraph into a line graph, the link prediction task is converted into a node classification task, which graph convolution progress can learn edge embeddings from graphs more effectively. Then we design a novel cross-scale contrastive learning framework on the line graph and the subgraph to maximize the mutual information of them, so that fuses the structure and feature information. The experimental results demonstrate that the proposed LGCL outperforms the state-of-the-art methods and has better performance on generalization and robustness.
comment: 37 pages
♻ ☆ Bayesian Optimization for Cascade-type Multi-stage Processes
Complex processes in science and engineering are often formulated as multistage decision-making problems. In this paper, we consider a type of multistage decision-making process called a cascade process. A cascade process is a multistage process in which the output of one stage is used as an input for the subsequent stage. When the cost of each stage is expensive, it is difficult to search for the optimal controllable parameters for each stage exhaustively. To address this problem, we formulate the optimization of the cascade process as an extension of the Bayesian optimization framework and propose two types of acquisition functions based on credible intervals and expected improvement. We investigate the theoretical properties of the proposed acquisition functions and demonstrate their effectiveness through numerical experiments. In addition, we consider an extension called suspension setting in which we are allowed to suspend the cascade process at the middle of the multistage decision-making process that often arises in practical problems. We apply the proposed method in a test problem involving a solar cell simulator, which was the motivation for this study.
comment: 70pages, 7 figures
♻ ☆ AirGNNs: Graph Neural Networks over the Air
Graph neural networks (GNNs) are information processing architectures that model representations from networked data and allow for decentralized implementation through localized communications. Existing GNN architectures often assume ideal communication links and ignore channel effects, such as fading and noise, leading to performance degradation in real-world implementation. This paper proposes graph neural networks over the air (AirGNNs), a novel GNN architecture that incorporates the communication model into the architecture. AirGNN modifies the graph convolutional operation that shifts graph signals over random communication graphs to take into account channel fading and noise when aggregating features from neighbors, thus, improving the architecture robustness to channel impairments during testing. We propose a stochastic gradient descent based method to train the AirGNN, and show that the training procedure converges to a stationary solution. Numerical simulations on decentralized source localization and multi-robot flocking corroborate theoretical findings and show superior performance of the AirGNN over wireless communication channels.
♻ ☆ Exploring the Limits of Indiscriminate Data Poisoning Attacks
Indiscriminate data poisoning attacks aim to decrease a model's test accuracy by injecting a small amount of corrupted training data. Despite significant interest, existing attacks remain relatively ineffective against modern machine learning (ML) architectures. In this work, we introduce the notion of model poisonability as a technical tool to explore the intrinsic limits of data poisoning attacks. We derive an easily computable threshold to establish and quantify a surprising phase transition phenomenon among popular ML models: data poisoning attacks become effective only when the poisoning ratio exceeds our threshold. Building on existing parameter corruption attacks and refining the Gradient Canceling attack, we perform extensive experiments to confirm our theoretical findings, test the predictability of our transition threshold, and significantly improve existing data poisoning baselines over a range of datasets and models. Our work highlights the critical role played by the poisoning ratio, and sheds new insights on existing empirical results, attacks and mitigation strategies in data poisoning.
♻ ☆ Bort: Towards Explainable Neural Networks with Bounded Orthogonal Constraint ICLR 2023
Deep learning has revolutionized human society, yet the black-box nature of deep neural networks hinders further application to reliability-demanded industries. In the attempt to unpack them, many works observe or impact internal variables to improve the comprehensibility and invertibility of the black-box models. However, existing methods rely on intuitive assumptions and lack mathematical guarantees. To bridge this gap, we introduce Bort, an optimizer for improving model explainability with boundedness and orthogonality constraints on model parameters, derived from the sufficient conditions of model comprehensibility and invertibility. We perform reconstruction and backtracking on the model representations optimized by Bort and observe a clear improvement in model explainability. Based on Bort, we are able to synthesize explainable adversarial samples without additional parameters and training. Surprisingly, we find Bort constantly improves the classification accuracy of various architectures including ResNet and DeiT on MNIST, CIFAR-10, and ImageNet. Code: https://github.com/zbr17/Bort.
comment: ICLR 2023 accepted
♻ ☆ Sampling Attacks on Meta Reinforcement Learning: A Minimax Formulation and Complexity Analysis
Meta reinforcement learning (meta RL), as a combination of meta-learning ideas and reinforcement learning (RL), enables the agent to adapt to different tasks using a few samples. However, this sampling-based adaptation also makes meta RL vulnerable to adversarial attacks. By manipulating the reward feedback from sampling processes in meta RL, an attacker can mislead the agent into building wrong knowledge from training experience, which deteriorates the agent's performance when dealing with different tasks after adaptation. This paper provides a game-theoretical underpinning for understanding this type of security risk. In particular, we formally define the sampling attack model as a Stackelberg game between the attacker and the agent, which yields a minimax formulation. It leads to two online attack schemes: Intermittent Attack and Persistent Attack, which enable the attacker to learn an optimal sampling attack, defined by an $\epsilon$-first-order stationary point, within $\mathcal{O}(\epsilon^{-2})$ iterations. These attack schemes freeride the learning progress concurrently without extra interactions with the environment. By corroborating the convergence results with numerical experiments, we observe that a minor effort of the attacker can significantly deteriorate the learning performance, and the minimax approach can also help robustify the meta RL algorithms.
comment: updates: github repo posted
♻ ☆ An ODE Model for Dynamic Matching in Heterogeneous Networks
We study the problem of dynamic matching in heterogeneous networks, where agents are subject to compatibility restrictions and stochastic arrival and departure times. In particular, we consider networks with one type of easy-to-match agents and multiple types of hard-to-match agents, each subject to its own compatibility constraints. Such a setting arises in many real-world applications, including kidney exchange programs and carpooling platforms. We introduce a novel approach to modeling dynamic matching by establishing the ordinary differential equation (ODE) model, which offers a new perspective for evaluating various matching algorithms. We study two algorithms, namely the Greedy and Patient Algorithms, where both algorithms prioritize matching compatible hard-to-match agents over easy-to-match agents in heterogeneous networks. Our results demonstrate the trade-off between the conflicting goals of matching agents quickly and optimally, offering insights into the design of real-world dynamic matching systems. We provide simulations and a real-world case study using data from the Organ Procurement and Transplantation Network to validate theoretical predictions.
♻ ☆ Commitment with Signaling under Double-sided Information Asymmetry
Information asymmetry in games enables players with the information advantage to manipulate others' beliefs by strategically revealing information to other players. This work considers a double-sided information asymmetry in a Bayesian Stackelberg game, where the leader's realized action, sampled from the mixed strategy commitment, is hidden from the follower. In contrast, the follower holds private information about his payoff. Given asymmetric information on both sides, an important question arises: \emph{Does the leader's information advantage outweigh the follower's?} We answer this question affirmatively in this work, where we demonstrate that by adequately designing a signaling device that reveals partial information regarding the leader's realized action to the follower, the leader can achieve a higher expected utility than that without signaling. Moreover, unlike previous works on the Bayesian Stackelberg game where mathematical programming tools are utilized, we interpret the leader's commitment as a probability measure over the belief space. Such a probabilistic language greatly simplifies the analysis and allows an indirect signaling scheme, leading to a geometric characterization of the equilibrium under the proposed game model.
comment: Working paper; 18 pages
♻ ☆ Inference and FDR Control for Simulated Markov Random Fields in High-dimension
This paper studies the consistency and statistical inference of simulated Markov random fields (MRFs) in a high dimensional background. Our estimators are based on the Markov chain Monte Carlo maximum likelihood estimation (MCMC-MLE) method, penalized by the Elastic-net. Under mild conditions that ensure a specific convergence rate of the MCMC method, the $\ell_{1}$ consistency of Elastic-net-penalized MCMC-MLE is obtained. We further propose a decorrelated score test based on the decorrelated score function and prove the asymptotic normality of the score function without the influence of many nuisance parameters under the assumption that it accelerates the convergence of the MCMC method. The one-step estimator for a single parameter of interest is constructed by linearizing the decorrelated score function to solve its root, and the normality and confidence interval for the true value, is established. We use different algorithms to control the false discovery rate (FDR) for multiple testing problems via classic p-values and novel e-values. Finally, we empirically validate the asymptotic theories and demonstrate both FDR control procedures in our article have good performance.
comment: 39 pages
♻ ☆ End-to-End Multi-Task Denoising for joint SDR and PESQ Optimization
Supervised learning based on a deep neural network recently has achieved substantial improvement on speech enhancement. Denoising networks learn mapping from noisy speech to clean one directly, or to a spectrum mask which is the ratio between clean and noisy spectra. In either case, the network is optimized by minimizing mean square error (MSE) between ground-truth labels and time-domain or spectrum output. However, existing schemes have either of two critical issues: spectrum and metric mismatches. The spectrum mismatch is a well known issue that any spectrum modification after short-time Fourier transform (STFT), in general, cannot be fully recovered after inverse short-time Fourier transform (ISTFT). The metric mismatch is that a conventional MSE metric is sub-optimal to maximize our target metrics, signal-to-distortion ratio (SDR) and perceptual evaluation of speech quality (PESQ). This paper presents a new end-to-end denoising framework with the goal of joint SDR and PESQ optimization. First, the network optimization is performed on the time-domain signals after ISTFT to avoid spectrum mismatch. Second, two loss functions which have improved correlations with SDR and PESQ metrics are proposed to minimize metric mismatch. The experimental result showed that the proposed denoising scheme significantly improved both SDR and PESQ performance over the existing methods.
♻ ☆ Fairness in Forecasting of Observations of Linear Dynamical Systems AAAI 2021
In machine learning, training data often capture the behaviour of multiple subgroups of some underlying human population. When the nature of training data for subgroups are not controlled carefully, under-representation bias arises. To counter this effect we introduce two natural notions of subgroup fairness and instantaneous fairness to address such under-representation bias in time-series forecasting problems. Here we show globally convergent methods for the fairness-constrained learning problems using hierarchies of convexifications of non-commutative polynomial optimisation problems. Our empirical results on a biased data set motivated by insurance applications and the well-known COMPAS data set demonstrate the efficacy of our methods. We also show that by exploiting sparsity in the convexifications, we can reduce the run time of our methods considerably.
comment: Journal version of Zhou et al. [arXiv:2006.07315, AAAI 2021]
♻ ☆ Disentangling representations in Restricted Boltzmann Machines without adversaries
A goal of unsupervised machine learning is to build representations of complex high-dimensional data, with simple relations to their properties. Such disentangled representations make easier to interpret the significant latent factors of variation in the data, as well as to generate new data with desirable features. Methods for disentangling representations often rely on an adversarial scheme, in which representations are tuned to avoid discriminators from being able to reconstruct information about the data properties (labels). Unfortunately adversarial training is generally difficult to implement in practice. Here we propose a simple, effective way of disentangling representations without any need to train adversarial discriminators, and apply our approach to Restricted Boltzmann Machines (RBM), one of the simplest representation-based generative models. Our approach relies on the introduction of adequate constraints on the weights during training, which allows us to concentrate information about labels on a small subset of latent variables. The effectiveness of the approach is illustrated with four examples: the CelebA dataset of facial images, the two-dimensional Ising model, the MNIST dataset of handwritten digits, and the taxonomy of protein families. In addition, we show how our framework allows for analytically computing the cost, in terms of log-likelihood of the data, associated to the disentanglement of their representations.
comment: Minor corrections. Accepted for publication in Physical Review X
♻ ☆ In Defense of the Unitary Scalarization for Deep Multi-Task Learning NeurIPS 2022
Recent multi-task learning research argues against unitary scalarization, where training simply minimizes the sum of the task losses. Several ad-hoc multi-task optimization algorithms have instead been proposed, inspired by various hypotheses about what makes multi-task settings difficult. The majority of these optimizers require per-task gradients, and introduce significant memory, runtime, and implementation overhead. We show that unitary scalarization, coupled with standard regularization and stabilization techniques from single-task learning, matches or improves upon the performance of complex multi-task optimizers in popular supervised and reinforcement learning settings. We then present an analysis suggesting that many specialized multi-task optimizers can be partly interpreted as forms of regularization, potentially explaining our surprising results. We believe our results call for a critical reevaluation of recent research in the area.
comment: NeurIPS 2022 camera-ready version, fixed training loss y axis scale
♻ ☆ On Fairness of Medical Image Classification with Multiple Sensitive Attributes via Learning Orthogonal Representations
Mitigating the discrimination of machine learning models has gained increasing attention in medical image analysis. However, rare works focus on fair treatments for patients with multiple sensitive demographic ones, which is a crucial yet challenging problem for real-world clinical applications. In this paper, we propose a novel method for fair representation learning with respect to multi-sensitive attributes. We pursue the independence between target and multi-sensitive representations by achieving orthogonality in the representation space. Concretely, we enforce the column space orthogonality by keeping target information on the complement of a low-rank sensitive space. Furthermore, in the row space, we encourage feature dimensions between target and sensitive representations to be orthogonal. The effectiveness of the proposed method is demonstrated with extensive experiments on the CheXpert dataset. To our best knowledge, this is the first work to mitigate unfairness with respect to multiple sensitive attributes in the field of medical imaging.
♻ ☆ Data-driven reduced-order modelling for blood flow simulations with geometry-informed snapshots
Computational fluid dynamics is a common tool in cardiovascular science and engineering to simulate, predict and study hemodynamics in arteries. However, owing to the complexity and scale of cardiovascular flow problems, the evaluation of the model could be computationally expensive, especially in those cases where a large number of evaluations are required, such as uncertainty quantification and design optimisation. In such scenarios, the model may have to be repeatedly evaluated due to the changes or distinctions of simulation domains. In this work, a data-driven surrogate model is proposed for the efficient prediction of blood flow simulations on similar but distinct domains. The proposed surrogate model leverages surface registration to parameterise those similar but distinct shapes and formulate corresponding hemodynamics information into geometry-informed snapshots by the diffeomorphism constructed between the reference domain and target domain. A non-intrusive reduced-order model for geometrical parameters is subsequently constructed using proper orthogonal decomposition, and a radial basis function interpolator is trained for predicting the reduced coefficients of the reduced-order model based on reduced coefficients of geometrical parameters of the shape. Two examples of blood flowing through a stenosis and a bifurcation are presented and analysed. The proposed surrogate model demonstrates its accuracy and efficiency in hemodynamics prediction and shows its potential application toward real-time simulation or uncertainty quantification for complex patient-specific scenarios.
♻ ☆ Active Learning and Bayesian Optimization: a Unified Perspective to Learn with a Goal
Both Bayesian optimization and active learning realize an adaptive sampling scheme to achieve a specific learning goal. However, while the two fields have seen an exponential growth in popularity in the past decade, their dualism has received relatively little attention. In this paper, we argue for an original unified perspective of Bayesian optimization and active learning based on the synergy between the principles driving the sampling policies. This symbiotic relationship is demonstrated through the substantial analogy between the infill criteria of Bayesian optimization and the learning criteria in active learning, and is formalized for the case of single information source and when multiple sources at different levels of fidelity are available. We further investigate the capabilities of each infill criteria both individually and in combination on a variety of analytical benchmark problems, to highlight benefits and limitations over mathematical properties that characterize real-world applications.
♻ ☆ Why do networks have inhibitory/negative connections?
Why do brains have inhibitory connections? Why do deep networks have negative weights? We believe representing functions is the primary role of both (i) the brain in natural intelligence, and (ii) deep networks in artificial intelligence. Our answer to why there are inhibitory/negative weights is: to learn more functions. We prove that, in the absence of negative weights, neural networks with non-decreasing activation functions are not universal approximators. While this may be an intuitive result to some, to the best of our knowledge, there is no formal theory, in either machine learning or neuroscience, that demonstrates why negative weights are crucial in the context of representation capacity. Further, we provide insights on the geometric properties of the representation space that non-negative deep networks cannot represent. We expect these insights will yield a deeper understanding of more sophisticated inductive priors imposed on the distribution of weights that lead to more efficient biological and machine learning.
♻ ☆ Global Concept-Based Interpretability for Graph Neural Networks via Neuron Analysis
Graph neural networks (GNNs) are highly effective on a variety of graph-related tasks; however, they lack interpretability and transparency. Current explainability approaches are typically local and treat GNNs as black-boxes. They do not look inside the model, inhibiting human trust in the model and explanations. Motivated by the ability of neurons to detect high-level semantic concepts in vision models, we perform a novel analysis on the behaviour of individual GNN neurons to answer questions about GNN interpretability, and propose new metrics for evaluating the interpretability of GNN neurons. We propose a novel approach for producing global explanations for GNNs using neuron-level concepts to enable practitioners to have a high-level view of the model. Specifically, (i) to the best of our knowledge, this is the first work which shows that GNN neurons act as concept detectors and have strong alignment with concepts formulated as logical compositions of node degree and neighbourhood properties; (ii) we quantitatively assess the importance of detected concepts, and identify a trade-off between training duration and neuron-level interpretability; (iii) we demonstrate that our global explainability approach has advantages over the current state-of-the-art -- we can disentangle the explanation into individual interpretable concepts backed by logical descriptions, which reduces potential for bias and improves user-friendliness.
comment: 9 pages, 5 figures
♻ ☆ Compositional optimization of quantum circuits for quantum kernels of support vector machines
While quantum machine learning (ML) has been proposed to be one of the most promising applications of quantum computing, how to build quantum ML models that outperform classical ML remains a major open question. Here, we demonstrate a Bayesian algorithm for constructing quantum kernels for support vector machines that adapts quantum gate sequences to data. The algorithm increases the complexity of quantum circuits incrementally by appending quantum gates selected with Bayesian information criterion as circuit selection metric and Bayesian optimization of the parameters of the locally optimal quantum circuits identified. The goal is to build quantum kernels for SVM that can solve classification problems with as little training data as possible. The performance of the resulting quantum models for the classification problems considered here significantly exceeds that of optimized classical models with conventional kernels.
♻ ☆ Local Convolutions Cause an Implicit Bias towards High Frequency Adversarial Examples
Adversarial Attacks are still a significant challenge for neural networks. Recent work has shown that adversarial perturbations typically contain high-frequency features, but the root cause of this phenomenon remains unknown. Inspired by theoretical work on linear full-width convolutional models, we hypothesize that the local (i.e. bounded-width) convolutional operations commonly used in current neural networks are implicitly biased to learn high frequency features, and that this is one of the root causes of high frequency adversarial examples. To test this hypothesis, we analyzed the impact of different choices of linear and nonlinear architectures on the implicit bias of the learned features and the adversarial perturbations, in both spatial and frequency domains. We find that the high-frequency adversarial perturbations are critically dependent on the convolution operation because the spatially-limited nature of local convolutions induces an implicit bias towards high frequency features. The explanation for the latter involves the Fourier Uncertainty Principle: a spatially-limited (local in the space domain) filter cannot also be frequency-limited (local in the frequency domain). Furthermore, using larger convolution kernel sizes or avoiding convolutions (e.g. by using Vision Transformers architecture) significantly reduces this high frequency bias, but not the overall susceptibility to attacks. Looking forward, our work strongly suggests that understanding and controlling the implicit bias of architectures will be essential for achieving adversarial robustness.
comment: 23 pages, 11 figures, 12 Tables
♻ ☆ Optimistic Whittle Index Policy: Online Learning for Restless Bandits AAAI 2023
Restless multi-armed bandits (RMABs) extend multi-armed bandits to allow for stateful arms, where the state of each arm evolves restlessly with different transitions depending on whether that arm is pulled. Solving RMABs requires information on transition dynamics, which are often unknown upfront. To plan in RMAB settings with unknown transitions, we propose the first online learning algorithm based on the Whittle index policy, using an upper confidence bound (UCB) approach to learn transition dynamics. Specifically, we estimate confidence bounds of the transition probabilities and formulate a bilinear program to compute optimistic Whittle indices using these estimates. Our algorithm, UCWhittle, achieves sublinear $O(H \sqrt{T \log T})$ frequentist regret to solve RMABs with unknown transitions in $T$ episodes with a constant horizon $H$. Empirically, we demonstrate that UCWhittle leverages the structure of RMABs and the Whittle index policy solution to achieve better performance than existing online learning baselines across three domains, including one constructed from a real-world maternal and childcare dataset.
comment: Accepted at AAAI 2023. 7 page paper, 2 page references, 9 page appendix. Code available. Proceedings of the Thirty-Seventh AAAI Conference on Artificial Intelligence (AAAI 2023)
Multimedia
☆ Multi-MEC Cooperation Based VR Video Transmission and Cache using K-Shortest Paths Optimization
In recent network architectures, multi-MEC cooperative caching has been introduced to reduce the transmission latency of VR videos, in which MEC servers' computing and caching capability are utilized to optimize the transmission process. However, many solutions that use the computing capability of MEC servers ignore the additional arithmetic power consumed by the codec process, thus making them infeasible. Besides, the minimum cache unit is usually the entire VR video, which makes caching inefficient. To address these challenges, we split VR videos into tile files for caching based on the current popular network architecture and provide a reliable transmission mechanism and an effective caching strategy. Since the number of different tile files N is too large, the current cooperative caching algorithms do not cope with such large-scale input data. We further analyze the problem and propose an optimized k-shortest paths (OKSP) algorithm with an upper bound time complexity of O((K * M + N) * M * logN)), and suitable for shortest paths with restricted number of edges, where K is the total number of tiles that all M MEC servers can cache in the collaboration domain. And we prove the OKSP algorithm can compute the caching scheme with the lowest average latency in any case, which means the solution given is the exact solution. The simulation results show that the OKSP algorithm has excellent speed for solving large-scale data and consistently outperforms other caching algorithms in the experiments.
♻ ☆ CaDM: Codec-aware Diffusion Modeling for Neural-enhanced Video Streaming
Recent years have witnessed the dramatic growth of Internet video traffic, where the video bitstreams are often compressed and delivered in low quality to fit the streamer's uplink bandwidth. To alleviate the quality degradation, it comes the rise of Neural-enhanced Video Streaming (NVS), which shows great prospects for recovering low-quality videos by mostly deploying neural super-resolution (SR) on the media server. Despite its benefit, we reveal that current mainstream works with SR enhancement have not achieved the desired rate-distortion trade-off between bitrate saving and quality restoration, due to: (1) overemphasizing the enhancement on the decoder side while omitting the co-design of encoder, (2) limited generative capacity to recover high-fidelity perceptual details, and (3) optimizing the compression-and-restoration pipeline from the resolution perspective solely, without considering color bit-depth. Aiming at overcoming these limitations, we are the first to conduct an encoder-decoder (i.e., codec) synergy by leveraging the inherent visual-generative property of diffusion models. Specifically, we present the Codec-aware Diffusion Modeling (CaDM), a novel NVS paradigm to significantly reduce streaming delivery bitrates while holding pretty higher restoration capacity over existing methods. First, CaDM improves the encoder's compression efficiency by simultaneously reducing resolution and color bit-depth of video frames. Second, CaDM empowers the decoder with high-quality enhancement by making the denoising diffusion restoration aware of encoder's resolution-color conditions. Evaluation on public cloud services with OpenMMLab benchmarks shows that CaDM effectively saves up to 5.12 - 21.44 times bitrates based on common video standards and achieves much better recovery quality (e.g., FID of 0.61) over state-of-the-art neural-enhancing methods.
♻ ☆ Video Question Answering Using CLIP-Guided Visual-Text Attention ICIP 2023
Cross-modal learning of video and text plays a key role in Video Question Answering (VideoQA). In this paper, we propose a visual-text attention mechanism to utilize the Contrastive Language-Image Pre-training (CLIP) trained on lots of general domain language-image pairs to guide the cross-modal learning for VideoQA. Specifically, we first extract video features using a TimeSformer and text features using a BERT from the target application domain, and utilize CLIP to extract a pair of visual-text features from the general-knowledge domain through the domain-specific learning. We then propose a Cross-domain Learning to extract the attention information between visual and linguistic features across the target domain and general domain. The set of CLIP-guided visual-text features are integrated to predict the answer. The proposed method is evaluated on MSVD-QA and MSRVTT-QA datasets, and outperforms state-of-the-art methods.
comment: Submitted to the 2023 IEEE International Conference on Image Processing (ICIP 2023)
Computation and Language
☆ A Hybrid Architecture for Out of Domain Intent Detection and Intent Discovery
Intent Detection is one of the tasks of the Natural Language Understanding (NLU) unit in task-oriented dialogue systems. Out of Scope (OOS) and Out of Domain (OOD) inputs may run these systems into a problem. On the other side, a labeled dataset is needed to train a model for Intent Detection in task-oriented dialogue systems. The creation of a labeled dataset is time-consuming and needs human resources. The purpose of this article is to address mentioned problems. The task of identifying OOD/OOS inputs is named OOD/OOS Intent Detection. Also, discovering new intents and pseudo-labeling of OOD inputs is well known by Intent Discovery. In OOD intent detection part, we make use of a Variational Autoencoder to distinguish between known and unknown intents independent of input data distribution. After that, an unsupervised clustering method is used to discover different unknown intents underlying OOD/OOS inputs. We also apply a non-linear dimensionality reduction on OOD/OOS representations to make distances between representations more meaning full for clustering. Our results show that the proposed model for both OOD/OOS Intent Detection and Intent Discovery achieves great results and passes baselines in English and Persian languages.
☆ Exploiting Asymmetry for Synthetic Training Data Generation: SynthIE and the Case of Information Extraction
Large language models (LLMs) show great potential for synthetic data generation. This work shows that useful data can be synthetically generated even for tasks that cannot be solved directly by the LLM: we show that, for problems with structured outputs, it is possible to prompt an LLM to perform the task in the opposite direction, to generate plausible text for the target structure. Leveraging the asymmetry in task difficulty makes it possible to produce large-scale, high-quality data for complex tasks. We demonstrate the effectiveness of this approach on closed information extraction, where collecting ground-truth data is challenging, and no satisfactory dataset exists to date. We synthetically generate a dataset of 1.8M data points, demonstrate its superior quality compared to existing datasets in a human evaluation and use it to finetune small models (220M and 770M parameters). The models we introduce, SynthIE, outperform existing baselines of comparable size with a substantial gap of 57 and 79 absolute points in micro and macro F1, respectively. Code, data, and models are available at https://github.com/epfl-dlab/SynthIE.
☆ Marpa and nullable symbols
The Marpa parser was intended to make the best results in the academic literature on Earley's algorithm available as a practical general parser. Earley-based parsers have had issues handling nullable symbols. Initially, we dealt with nullable symbols by following the approach in Aycock and Horspool's 2002 paper. This paper reports our experience with the approach in that paper, and the approach to handling nullables that we settled on in reaction to that experience.
☆ CroCoSum: A Benchmark Dataset for Cross-Lingual Code-Switched Summarization
Cross-lingual summarization (CLS) has attracted increasing interest in recent years due to the availability of large-scale web-mined datasets and the advancements of multilingual language models. However, given the rareness of naturally occurring CLS resources, the majority of datasets are forced to rely on translation which can contain overly literal artifacts. This restricts our ability to observe naturally occurring CLS pairs that capture organic diction, including instances of code-switching. This alteration between languages in mid-message is a common phenomenon in multilingual settings yet has been largely overlooked in cross-lingual contexts due to data scarcity. To address this gap, we introduce CroCoSum, a dataset of cross-lingual code-switched summarization of technology news. It consists of over 24,000 English source articles and 18,000 human-curated Chinese news summaries, with more than 92% of the summaries containing code-switched phrases. For reference, we evaluate the performance of existing approaches including pipeline, end-to-end, and zero-shot methods. We show that leveraging existing resources as a pretraining step does not improve performance on CroCoSum, indicating the limited generalizability of existing resources. Finally, we discuss the challenges of evaluating cross-lingual summarizers on code-switched generation through qualitative error analyses. Our collection and code can be accessed at https://github.com/RosenZhang/CroCoSum.
comment: Work in Progress
☆ Visual Abstraction and Reasoning through Language
While Artificial Intelligence (AI) models have achieved human or even superhuman performance in narrowly defined applications, they still struggle to show signs of broader and more flexible intelligence. The Abstraction and Reasoning Corpus (ARC), introduced by Fran\c{c}ois Chollet, aims to assess how close AI systems are to human-like cognitive abilities. Most current approaches rely on carefully handcrafted domain-specific languages (DSLs), which are used to brute-force solutions to the tasks present in ARC. In this work, we propose a general framework for solving ARC based on natural language descriptions of the tasks. While not yet beating state-of-the-art DSL models on ARC, we demonstrate the immense potential of our approach hinted at by the ability to solve previously unsolved tasks.
comment: The first two authors have contributed equally to this work
☆ Describe me an Aucklet: Generating Grounded Perceptual Category Descriptions
Human language users can generate descriptions of perceptual concepts beyond instance-level representations and also use such descriptions to learn provisional class-level representations. However, the ability of computational models to learn and operate with class representations is under-investigated in the language-and-vision field. In this paper, we train separate neural networks to generate and interpret class-level descriptions. We then use the zero-shot classification performance of the interpretation model as a measure of communicative success and class-level conceptual grounding. We investigate the performance of prototype- and exemplar-based neural representations grounded category description. Finally, we show that communicative success reveals performance issues in the generation model that are not captured by traditional intrinsic NLG evaluation metrics, and argue that these issues can be traced to a failure to properly ground language in vision at the class level. We observe that the interpretation model performs better with descriptions that are low in diversity on the class level, possibly indicating a strong reliance on frequently occurring features.
☆ Is ChatGPT a Good NLG Evaluator? A Preliminary Study
Recently, the emergence of ChatGPT has attracted wide attention from the computational linguistics community. Many prior studies have shown that ChatGPT achieves remarkable performance on various NLP tasks in terms of automatic evaluation metrics. However, the ability of ChatGPT to serve as an evaluation metric is still underexplored. Considering assessing the quality of NLG models is an arduous task and previous statistical metrics notoriously show their poor correlation with human judgments, we wonder whether ChatGPT is a good NLG evaluation metric. In this report, we provide a preliminary meta-evaluation on ChatGPT to show its reliability as an NLG metric. In detail, we regard ChatGPT as a human evaluator and give task-specific (e.g., summarization) and aspect-specific (e.g., relevance) instruction to prompt ChatGPT to score the generation of NLG models. We conduct experiments on three widely-used NLG meta-evaluation datasets (including summarization, story generation and data-to-text tasks). Experimental results show that compared with previous automatic metrics, ChatGPT achieves state-of-the-art or competitive correlation with golden human judgments. We hope our preliminary study could prompt the emergence of a general-purposed reliable NLG metric.
comment: Technical Report, 8 pages
☆ ELODIN: Naming Concepts in Embedding Spaces
Despite recent advancements, the field of text-to-image synthesis still suffers from lack of fine-grained control. Using only text, it remains challenging to deal with issues such as concept coherence and concept contamination. We propose a method to enhance control by generating specific concepts that can be reused throughout multiple images, effectively expanding natural language with new words that can be combined much like a painter's palette. Unlike previous contributions, our method does not copy visuals from input data and can generate concepts through text alone. We perform a set of comparisons that finds our method to be a significant improvement over text-only prompts.
☆ GATE: A Challenge Set for Gender-Ambiguous Translation Examples
Although recent years have brought significant progress in improving translation of unambiguously gendered sentences, translation of ambiguously gendered input remains relatively unexplored. When source gender is ambiguous, machine translation models typically default to stereotypical gender roles, perpetuating harmful bias. Recent work has led to the development of "gender rewriters" that generate alternative gender translations on such ambiguous inputs, but such systems are plagued by poor linguistic coverage. To encourage better performance on this task we present and release GATE, a linguistically diverse corpus of gender-ambiguous source sentences along with multiple alternative target language translations. We also provide tools for evaluation and system analysis when using GATE and use them to evaluate our translation rewriter system.
☆ ChatGPT: Beginning of an End of Manual Annotation? Use Case of Automatic Genre Identification
ChatGPT has shown strong capabilities in natural language generation tasks, which naturally leads researchers to explore where its abilities end. In this paper, we examine whether ChatGPT can be used for zero-shot text classification, more specifically, automatic genre identification. We compare ChatGPT with a multilingual XLM-RoBERTa language model that was fine-tuned on datasets, manually annotated with genres. The models are compared on test sets in two languages: English and Slovenian. Results show that ChatGPT outperforms the fine-tuned model when applied to the dataset which was not seen before by either of the models. Even when applied on Slovenian language as an under-resourced language, ChatGPT's performance is no worse than when applied to English. However, if the model is fully prompted in Slovenian, the performance drops significantly, showing the current limitations of ChatGPT usage on smaller languages. The presented results lead us to questioning whether this is the beginning of an end of laborious manual annotation campaigns even for smaller languages, such as Slovenian.
☆ A Meta-Evaluation of Faithfulness Metrics for Long-Form Hospital-Course Summarization
Long-form clinical summarization of hospital admissions has real-world significance because of its potential to help both clinicians and patients. The faithfulness of summaries is critical to their safe usage in clinical settings. To better understand the limitations of abstractive systems, as well as the suitability of existing evaluation metrics, we benchmark faithfulness metrics against fine-grained human annotations for model-generated summaries of a patient's Brief Hospital Course. We create a corpus of patient hospital admissions and summaries for a cohort of HIV patients, each with complex medical histories. Annotators are presented with summaries and source notes, and asked to categorize manually highlighted summary elements (clinical entities like conditions and medications as well as actions like "following up") into one of three categories: ``Incorrect,'' ``Missing,'' and ``Not in Notes.'' We meta-evaluate a broad set of proposed faithfulness metrics and, across metrics, explore the importance of domain adaptation (e.g. the impact of in-domain pre-training and metric fine-tuning), the use of source-summary alignments, and the effects of distilling a single metric from an ensemble of pre-existing metrics. Off-the-shelf metrics with no exposure to clinical text correlate well yet overly rely on summary extractiveness. As a practical guide to long-form clinical narrative summarization, we find that most metrics correlate best to human judgments when provided with one summary sentence at a time and a minimal set of relevant source context.
comment: Preprint
☆ Speak Foreign Languages with Your Own Voice: Cross-Lingual Neural Codec Language Modeling
We propose a cross-lingual neural codec language model, VALL-E X, for cross-lingual speech synthesis. Specifically, we extend VALL-E and train a multi-lingual conditional codec language model to predict the acoustic token sequences of the target language speech by using both the source language speech and the target language text as prompts. VALL-E X inherits strong in-context learning capabilities and can be applied for zero-shot cross-lingual text-to-speech synthesis and zero-shot speech-to-speech translation tasks. Experimental results show that it can generate high-quality speech in the target language via just one speech utterance in the source language as a prompt while preserving the unseen speaker's voice, emotion, and acoustic environment. Moreover, VALL-E X effectively alleviates the foreign accent problems, which can be controlled by a language ID. Audio samples are available at \url{https://aka.ms/vallex}.
comment: We encourage readers to listen to the audio samples on our demo page: \url{https://aka.ms/vallex}
☆ The BigScience ROOTS Corpus: A 1.6TB Composite Multilingual Dataset NeurIPS 2022
As language models grow ever larger, the need for large-scale high-quality text datasets has never been more pressing, especially in multilingual settings. The BigScience workshop, a 1-year international and multidisciplinary initiative, was formed with the goal of researching and training large language models as a values-driven undertaking, putting issues of ethics, harm, and governance in the foreground. This paper documents the data creation and curation efforts undertaken by BigScience to assemble the Responsible Open-science Open-collaboration Text Sources (ROOTS) corpus, a 1.6TB dataset spanning 59 languages that was used to train the 176-billion-parameter BigScience Large Open-science Open-access Multilingual (BLOOM) language model. We further release a large initial subset of the corpus and analyses thereof, and hope to empower large-scale monolingual and multilingual modeling projects with both the data and the processing tools, as well as stimulate research around this large multilingual corpus.
comment: NeurIPS 2022, Datasets and Benchmarks Track
☆ Document-level Relation Extraction with Cross-sentence Reasoning Graph PAKDD 2023
Relation extraction (RE) has recently moved from the sentence-level to document-level, which requires aggregating document information and using entities and mentions for reasoning. Existing works put entity nodes and mention nodes with similar representations in a document-level graph, whose complex edges may incur redundant information. Furthermore, existing studies only focus on entity-level reasoning paths without considering global interactions among entities cross-sentence. To these ends, we propose a novel document-level RE model with a GRaph information Aggregation and Cross-sentence Reasoning network (GRACR). Specifically, a simplified document-level graph is constructed to model the semantic information of all mentions and sentences in a document, and an entity-level graph is designed to explore relations of long-distance cross-sentence entity pairs. Experimental results show that GRACR achieves excellent performance on two public datasets of document-level RE. It is especially effective in extracting potential relations of cross-sentence entity pairs. Our code is available at https://github.com/UESTC-LHF/GRACR.
comment: This paper is accepted by PAKDD 2023
☆ Larger language models do in-context learning differently
We study how in-context learning (ICL) in language models is affected by semantic priors versus input-label mappings. We investigate two setups-ICL with flipped labels and ICL with semantically-unrelated labels-across various model families (GPT-3, InstructGPT, Codex, PaLM, and Flan-PaLM). First, experiments on ICL with flipped labels show that overriding semantic priors is an emergent ability of model scale. While small language models ignore flipped labels presented in-context and thus rely primarily on semantic priors from pretraining, large models can override semantic priors when presented with in-context exemplars that contradict priors, despite the stronger semantic priors that larger models may hold. We next study semantically-unrelated label ICL (SUL-ICL), in which labels are semantically unrelated to their inputs (e.g., foo/bar instead of negative/positive), thereby forcing language models to learn the input-label mappings shown in in-context exemplars in order to perform the task. The ability to do SUL-ICL also emerges primarily with scale, and large-enough language models can even perform linear classification in a SUL-ICL setting. Finally, we evaluate instruction-tuned models and find that instruction tuning strengthens both the use of semantic priors and the capacity to learn input-label mappings, but more of the former.
☆ A Challenging Benchmark for Low-Resource Learning
With promising yet saturated results in high-resource settings, low-resource datasets have gradually become popular benchmarks for evaluating the learning ability of advanced neural networks (e.g., BigBench, superGLUE). Some models even surpass humans according to benchmark test results. However, we find that there exists a set of hard examples in low-resource settings that challenge neural networks but are not well evaluated, which causes over-estimated performance. We first give a theoretical analysis on which factors bring the difficulty of low-resource learning. It then motivate us to propose a challenging benchmark hardBench to better evaluate the learning ability, which covers 11 datasets, including 3 computer vision (CV) datasets and 8 natural language process (NLP) datasets. Experiments on a wide range of models show that neural networks, even pre-trained language models, have sharp performance drops on our benchmark, demonstrating the effectiveness on evaluating the weaknesses of neural networks. On NLP tasks, we surprisingly find that despite better results on traditional low-resource benchmarks, pre-trained networks, does not show performance improvements on our benchmarks. These results demonstrate that there are still a large robustness gap between existing models and human-level performance.
☆ Exploring the Feasibility of ChatGPT for Event Extraction
Event extraction is a fundamental task in natural language processing that involves identifying and extracting information about events mentioned in text. However, it is a challenging task due to the lack of annotated data, which is expensive and time-consuming to obtain. The emergence of large language models (LLMs) such as ChatGPT provides an opportunity to solve language tasks with simple prompts without the need for task-specific datasets and fine-tuning. While ChatGPT has demonstrated impressive results in tasks like machine translation, text summarization, and question answering, it presents challenges when used for complex tasks like event extraction. Unlike other tasks, event extraction requires the model to be provided with a complex set of instructions defining all event types and their schemas. To explore the feasibility of ChatGPT for event extraction and the challenges it poses, we conducted a series of experiments. Our results show that ChatGPT has, on average, only 51.04% of the performance of a task-specific model such as EEQA in long-tail and complex scenarios. Our usability testing experiments indicate that ChatGPT is not robust enough, and continuous refinement of the prompt does not lead to stable performance improvements, which can result in a poor user experience. Besides, ChatGPT is highly sensitive to different prompt styles.
☆ Preparing the Vuk'uzenzele and ZA-gov-multilingual South African multilingual corpora
This paper introduces two multilingual government themed corpora in various South African languages. The corpora were collected by gathering the South African Government newspaper (Vuk'uzenzele), as well as South African government speeches (ZA-gov-multilingual), that are translated into all 11 South African official languages. The corpora can be used for a myriad of downstream NLP tasks. The corpora were created to allow researchers to study the language used in South African government publications, with a focus on understanding how South African government officials communicate with their constituents. In this paper we highlight the process of gathering, cleaning and making available the corpora. We create parallel sentence corpora for Neural Machine Translation (NMT) tasks using Language-Agnostic Sentence Representations (LASER) embeddings. With these aligned sentences we then provide NMT benchmarks for 9 indigenous languages by fine-tuning a massively multilingual pre-trained language model. \end{abstra
comment: Under Review
☆ Universal resources for quantum computing
Unravelling the source of quantum computing power has been a major goal in the field of quantum information science. In recent years, the quantum resource theory (QRT) has been established to characterize various quantum resources, yet their roles in quantum computing tasks still require investigation. The so-called universal quantum computing model (UQCM), e.g., the circuit model, has been the main framework to guide the design of quantum algorithms, creation of real quantum computers etc. In this work, we combine the study of UQCM together with QRT. We find on one hand, using QRT can provide a resource-theoretic characterization of a UQCM, the relation among models and inspire new ones, and on the other hand, using UQCM offers a framework to apply resources, study relation among resources and classify them. We develop the theory of universal resources in the setting of UQCM, and find a rich spectrum of UQCMs and the corresponding universal resources. Depending on a hierarchical structure of resource theories, we find models can be classified into families. In this work, we study three natural families of UQCMs in details: the amplitude family, the quasi-probability family, and the Hamiltonian family. They include some well known models, like the measurement-based model and adiabatic model, and also inspire new models such as the contextual model we introduce. Each family contains at least a triplet of models, and such a succinct structure of families of UQCMs offers a unifying picture to investigate resources and design models. It also provides a rigorous framework to resolve puzzles, such as the role of entanglement vs. interference, and unravel resource-theoretic features of quantum algorithms.
☆ Classifying Text-Based Conspiracy Tweets related to COVID-19 using Contextualized Word Embeddings
The FakeNews task in MediaEval 2022 investigates the challenge of finding accurate and high-performance models for the classification of conspiracy tweets related to COVID-19. In this paper, we used BERT, ELMO, and their combination for feature extraction and RandomForest as classifier. The results show that ELMO performs slightly better than BERT, however their combination at feature level reduces the performance.
comment: Published in Multimedia Benchmark Workshop 2022, Bergen, Norway and Online, 12-13 January 2023: https://2022.multimediaeval.com/
☆ Stylometric Detection of AI-Generated Text in Twitter Timelines
Recent advancements in pre-trained language models have enabled convenient methods for generating human-like text at a large scale. Though these generation capabilities hold great potential for breakthrough applications, it can also be a tool for an adversary to generate misinformation. In particular, social media platforms like Twitter are highly susceptible to AI-generated misinformation. A potential threat scenario is when an adversary hijacks a credible user account and incorporates a natural language generator to generate misinformation. Such threats necessitate automated detectors for AI-generated tweets in a given user's Twitter timeline. However, tweets are inherently short, thus making it difficult for current state-of-the-art pre-trained language model-based detectors to accurately detect at what point the AI starts to generate tweets in a given Twitter timeline. In this paper, we present a novel algorithm using stylometric signals to aid detecting AI-generated tweets. We propose models corresponding to quantifying stylistic changes in human and AI tweets in two related tasks: Task 1 - discriminate between human and AI-generated tweets, and Task 2 - detect if and when an AI starts to generate tweets in a given Twitter timeline. Our extensive experiments demonstrate that the stylometric features are effective in augmenting the state-of-the-art AI-generated text detectors.
☆ CoTEVer: Chain of Thought Prompting Annotation Toolkit for Explanation Verification EACL 2023
Chain-of-thought (CoT) prompting enables large language models (LLMs) to solve complex reasoning tasks by generating an explanation before the final prediction. Despite it's promising ability, a critical downside of CoT prompting is that the performance is greatly affected by the factuality of the generated explanation. To improve the correctness of the explanations, fine-tuning language models with explanation data is needed. However, there exists only a few datasets that can be used for such approaches, and no data collection tool for building them. Thus, we introduce CoTEVer, a tool-kit for annotating the factual correctness of generated explanations and collecting revision data of wrong explanations. Furthermore, we suggest several use cases where the data collected with CoTEVer can be utilized for enhancing the faithfulness of explanations. Our toolkit is publicly available at https://github.com/SeungoneKim/CoTEVer.
comment: Accepted at EACL 2023 Demo
☆ Towards Interpretable and Efficient Automatic Reference-Based Summarization Evaluation
Interpretability and efficiency are two important considerations for the adoption of neural automatic metrics. In this work, we develop strong-performing automatic metrics for reference-based summarization evaluation, based on a two-stage evaluation pipeline that first extracts basic information units from one text sequence and then checks the extracted units in another sequence. The metrics we developed include two-stage metrics that can provide high interpretability at both the fine-grained unit level and summary level, and one-stage metrics that achieve a balance between efficiency and interoperability. We make the developed tools publicly available through a Python package and GitHub.
comment: GitHub Repo: https://github.com/Yale-LILY/AutoACU
☆ Adaptive Knowledge Distillation between Text and Speech Pre-trained Models
Learning on a massive amount of speech corpus leads to the recent success of many self-supervised speech models. With knowledge distillation, these models may also benefit from the knowledge encoded by language models that are pre-trained on rich sources of texts. The distillation process, however, is challenging due to the modal disparity between textual and speech embedding spaces. This paper studies metric-based distillation to align the embedding space of text and speech with only a small amount of data without modifying the model structure. Since the semantic and granularity gap between text and speech has been omitted in literature, which impairs the distillation, we propose the Prior-informed Adaptive knowledge Distillation (PAD) that adaptively leverages text/speech units of variable granularity and prior distributions to achieve better global and local alignments between text and speech pre-trained models. We evaluate on three spoken language understanding benchmarks to show that PAD is more effective in transferring linguistic knowledge than other metric-based distillation approaches.
☆ ADELT: Transpilation Between Deep Learning Frameworks
We propose Adversarial DEep Learning Transpiler (ADELT) for source-to-source transpilation between deep learning frameworks. Unlike prior approaches, we decouple the transpilation of code skeletons and the mapping of API keywords (an API function name or a parameter name). ADELT transpile code skeletons using few-shot prompting on big language models. Based on contextual embeddings extracted by a BERT for code, we train aligned API embeddings in a domain-adversarial setup, upon which we generate a dictionary for keyword translation. The model is trained on our unlabeled DL corpus from web crawl data, without using any hand-crafted rules and parallel data. Our method outperforms state-of-the-art transpilers on multiple transpilation pairs including PyTorch-Keras and PyTorch-MXNet by 15.9pts and 12.0pts in exact match scores respectively.
comment: 23 pages
☆ Do Prosody Transfer Models Transfer Prosody? ICASSP 2023
Some recent models for Text-to-Speech synthesis aim to transfer the prosody of a reference utterance to the generated target synthetic speech. This is done by using a learned embedding of the reference utterance, which is used to condition speech generation. During training, the reference utterance is identical to the target utterance. Yet, during synthesis, these models are often used to transfer prosody from a reference that differs from the text or speaker being synthesized. To address this inconsistency, we propose to use a different, but prosodically-related, utterance during training too. We believe this should encourage the model to learn to transfer only those characteristics that the reference and target have in common. If prosody transfer methods do indeed transfer prosody they should be able to be trained in the way we propose. However, results show that a model trained under these conditions performs significantly worse than one trained using the target utterance as a reference. To explain this, we hypothesize that prosody transfer models do not learn a transferable representation of prosody, but rather an utterance-level representation which is highly dependent on both the reference speaker and reference text.
comment: Accepted in ICASSP 2023, 5 pages, 2 figures, 3 tables
☆ How Do Transformers Learn Topic Structure: Towards a Mechanistic Understanding
While the successes of transformers across many domains are indisputable, accurate understanding of the learning mechanics is still largely lacking. Their capabilities have been probed on benchmarks which include a variety of structured and reasoning tasks -- but mathematical understanding is lagging substantially behind. Recent lines of work have begun studying representational aspects of this question: that is, the size/depth/complexity of attention-based networks to perform certain tasks. However, there is no guarantee the learning dynamics will converge to the constructions proposed. In our paper, we provide fine-grained mechanistic understanding of how transformers learn "semantic structure", understood as capturing co-occurrence structure of words. Precisely, we show, through a combination of experiments on synthetic data modeled by Latent Dirichlet Allocation (LDA), Wikipedia data, and mathematical analysis that the embedding layer and the self-attention layer encode the topical structure. In the former case, this manifests as higher average inner product of embeddings between same-topic words. In the latter, it manifests as higher average pairwise attention between same-topic words. The mathematical results involve several assumptions to make the analysis tractable, which we verify on data, and might be of independent interest as well.
☆ A Comprehensive Survey of AI-Generated Content (AIGC): A History of Generative AI from GAN to ChatGPT
Recently, ChatGPT, along with DALL-E-2 and Codex,has been gaining significant attention from society. As a result, many individuals have become interested in related resources and are seeking to uncover the background and secrets behind its impressive performance. In fact, ChatGPT and other Generative AI (GAI) techniques belong to the category of Artificial Intelligence Generated Content (AIGC), which involves the creation of digital content, such as images, music, and natural language, through AI models. The goal of AIGC is to make the content creation process more efficient and accessible, allowing for the production of high-quality content at a faster pace. AIGC is achieved by extracting and understanding intent information from instructions provided by human, and generating the content according to its knowledge and the intent information. In recent years, large-scale models have become increasingly important in AIGC as they provide better intent extraction and thus, improved generation results. With the growth of data and the size of the models, the distribution that the model can learn becomes more comprehensive and closer to reality, leading to more realistic and high-quality content generation. This survey provides a comprehensive review on the history of generative models, and basic components, recent advances in AIGC from unimodal interaction and multimodal interaction. From the perspective of unimodality, we introduce the generation tasks and relative models of text and image. From the perspective of multimodality, we introduce the cross-application between the modalities mentioned above. Finally, we discuss the existing open problems and future challenges in AIGC.
comment: 44 pages, 15 figures
☆ SemEval-2023 Task 10: Explainable Detection of Online Sexism SemEval-2023
Online sexism is a widespread and harmful phenomenon. Automated tools can assist the detection of sexism at scale. Binary detection, however, disregards the diversity of sexist content, and fails to provide clear explanations for why something is sexist. To address this issue, we introduce SemEval Task 10 on the Explainable Detection of Online Sexism (EDOS). We make three main contributions: i) a novel hierarchical taxonomy of sexist content, which includes granular vectors of sexism to aid explainability; ii) a new dataset of 20,000 social media comments with fine-grained labels, along with larger unlabelled datasets for model adaptation; and iii) baseline models as well as an analysis of the methods, results and errors for participant submissions to our task.
comment: SemEval-2023 Task 10 (ACL 2023)
☆ Gradient-Free Structured Pruning with Unlabeled Data
Large Language Models (LLMs) have achieved great success in solving difficult tasks across many domains, but such success comes with a high computation cost, and inference latency. As developers and third parties customize these models, the need to provide efficient inference has increased. Many efforts have attempted to reduce inference cost through model compression techniques such as pruning and distillation. However, these techniques either require labeled data, or are time-consuming as they require the compressed model to be retrained to regain accuracy. In this paper, we propose a gradient-free structured pruning framework that uses only unlabeled data. An evaluation on the GLUE and SQuAD benchmarks using BERT$_{BASE}$ and DistilBERT illustrates the effectiveness of the proposed approach. By only using the weights of the pre-trained model and unlabeled data, in a matter of a few minutes on a single GPU, up to 40% of the original FLOP count can be reduced with less than a 4% accuracy loss across all tasks considered.
☆ Can large language models build causal graphs?
Building causal graphs can be a laborious process. To ensure all relevant causal pathways have been captured, researchers often have to discuss with clinicians and experts while also reviewing extensive relevant medical literature. By encoding common and medical knowledge, large language models (LLMs) represent an opportunity to ease this process by automatically scoring edges (i.e., connections between two variables) in potential graphs. LLMs however have been shown to be brittle to the choice of probing words, context, and prompts that the user employs. In this work, we evaluate if LLMs can be a useful tool in complementing causal graph development.
☆ Extracting Accurate Materials Data from Research Papers with Conversational Language Models and Prompt Engineering -- Example of ChatGPT
There has been a growing effort to replace hand extraction of data from research papers with automated data extraction based on natural language processing (NLP), language models (LMs), and recently, large language models (LLMs). Although these methods enable efficient extraction of data from large sets of research papers, they require a significant amount of up-front effort, expertise, and coding. In this work we propose the ChatExtract method that can fully automate very accurate data extraction with essentially no initial effort or background using an advanced conversational LLM (or AI). ChatExtract consists of a set of engineered prompts applied to a conversational LLM that both identify sentences with data, extract data, and assure its correctness through a series of follow-up questions. These follow-up questions address a critical challenge associated with LLMs - their tendency to provide factually inaccurate responses. ChatExtract can be applied with any conversational LLMs and yields very high quality data extraction. In tests on materials data we find precision and recall both over 90% from the best conversational LLMs, likely rivaling or exceeding human accuracy in many cases. We demonstrate that the exceptional performance is enabled by the information retention in a conversational model combined with purposeful redundancy and introducing uncertainty through follow-up prompts. These results suggest that approaches similar to ChatExtract, due to their simplicity, transferability and accuracy are likely to replace other methods of data extraction in the near future.
comment: 7 pages, 2 figures, 1 table
☆ Automatically Summarizing Evidence from Clinical Trials: A Prototype Highlighting Current Challenges
We present TrialsSummarizer, a system that aims to automatically summarize evidence presented in the set of randomized controlled trials most relevant to a given query. Building on prior work, the system retrieves trial publications matching a query specifying a combination of condition, intervention(s), and outcome(s), and ranks these according to sample size and estimated study quality. The top-k such studies are passed through a neural multi-document summarization system, yielding a synopsis of these trials. We consider two architectures: A standard sequence-to-sequence model based on BART and a multi-headed architecture intended to provide greater transparency to end-users. Both models produce fluent and relevant summaries of evidence retrieved for queries, but their tendency to introduce unsupported statements render them inappropriate for use in this domain at present. The proposed architecture may help users verify outputs allowing users to trace generated tokens back to inputs.
☆ Making a Computational Attorney SDM'23
This "blue sky idea" paper outlines the opportunities and challenges in data mining and machine learning involving making a computational attorney -- an intelligent software agent capable of helping human lawyers with a wide range of complex high-level legal tasks such as drafting legal briefs for the prosecution or defense in court. In particular, we discuss what a ChatGPT-like Large Legal Language Model (L$^3$M) can and cannot do today, which will inspire researchers with promising short-term and long-term research objectives.
comment: To be published in the Proceedings of the 2023 SIAM International Conference on Data Mining (SDM'23)
☆ Disambiguation of Company names via Deep Recurrent Networks
Name Entity Disambiguation is the Natural Language Processing task of identifying textual records corresponding to the same Named Entity, i.e. real-world entities represented as a list of attributes (names, places, organisations, etc.). In this work, we face the task of disambiguating companies on the basis of their written names. We propose a Siamese LSTM Network approach to extract -- via supervised learning -- an embedding of company name strings in a (relatively) low dimensional vector space and use this representation to identify pairs of company names that actually represent the same company (i.e. the same Entity). Given that the manual labelling of string pairs is a rather onerous task, we analyse how an Active Learning approach to prioritise the samples to be labelled leads to a more efficient overall learning pipeline. With empirical investigations, we show that our proposed Siamese Network outperforms several benchmark approaches based on standard string matching algorithms when enough labelled data are available. Moreover, we show that Active Learning prioritisation is indeed helpful when labelling resources are limited, and let the learning models reach the out-of-sample performance saturation with less labelled data with respect to standard (random) data labelling approaches.
comment: submitted to Elsevier. 26 pages, 6 figures, 4 tables
☆ German BERT Model for Legal Named Entity Recognition
The use of BERT, one of the most popular language models, has led to improvements in many Natural Language Processing (NLP) tasks. One such task is Named Entity Recognition (NER) i.e. automatic identification of named entities such as location, person, organization, etc. from a given text. It is also an important base step for many NLP tasks such as information extraction and argumentation mining. Even though there is much research done on NER using BERT and other popular language models, the same is not explored in detail when it comes to Legal NLP or Legal Tech. Legal NLP applies various NLP techniques such as sentence similarity or NER specifically on legal data. There are only a handful of models for NER tasks using BERT language models, however, none of these are aimed at legal documents in German. In this paper, we fine-tune a popular BERT language model trained on German data (German BERT) on a Legal Entity Recognition (LER) dataset. To make sure our model is not overfitting, we performed a stratified 10-fold cross-validation. The results we achieve by fine-tuning German BERT on the LER dataset outperform the BiLSTM-CRF+ model used by the authors of the same LER dataset. Finally, we make the model openly available via HuggingFace.
comment: Presented at ICAART 2023
♻ ☆ Self-Consistency Improves Chain of Thought Reasoning in Language Models ICLR 2023
Chain-of-thought prompting combined with pre-trained large language models has achieved encouraging results on complex reasoning tasks. In this paper, we propose a new decoding strategy, self-consistency, to replace the naive greedy decoding used in chain-of-thought prompting. It first samples a diverse set of reasoning paths instead of only taking the greedy one, and then selects the most consistent answer by marginalizing out the sampled reasoning paths. Self-consistency leverages the intuition that a complex reasoning problem typically admits multiple different ways of thinking leading to its unique correct answer. Our extensive empirical evaluation shows that self-consistency boosts the performance of chain-of-thought prompting with a striking margin on a range of popular arithmetic and commonsense reasoning benchmarks, including GSM8K (+17.9%), SVAMP (+11.0%), AQuA (+12.2%), StrategyQA (+6.4%) and ARC-challenge (+3.9%).
comment: Published at ICLR 2023. V2: added PaLM results; V3: added UL2 results; V4: camera ready version at ICLR 2023
♻ ☆ MuAViC: A Multilingual Audio-Visual Corpus for Robust Speech Recognition and Robust Speech-to-Text Translation
We introduce MuAViC, a multilingual audio-visual corpus for robust speech recognition and robust speech-to-text translation providing 1200 hours of audio-visual speech in 9 languages. It is fully transcribed and covers 6 English-to-X translation as well as 6 X-to-English translation directions. To the best of our knowledge, this is the first open benchmark for audio-visual speech-to-text translation and the largest open benchmark for multilingual audio-visual speech recognition. Our baseline results show that MuAViC is effective for building noise-robust speech recognition and translation models. We make the corpus available at https://github.com/facebookresearch/muavic.
♻ ☆ LambdaKG: A Library for Pre-trained Language Model-Based Knowledge Graph Embeddings
Knowledge Graphs (KGs) often have two characteristics: heterogeneous graph structure and text-rich entity/relation information. Text-based KG embeddings can represent entities by encoding descriptions with pre-trained language models, but no open-sourced library is specifically designed for KGs with PLMs at present. In this paper, we present LambdaKG, a library for KGE that equips with many pre-trained language models (e.g., BERT, BART, T5, GPT-3), and supports various tasks (e.g., knowledge graph completion, question answering, recommendation, and knowledge probing). LambdaKG is publicly open-sourced at https://github.com/zjunlp/PromptKG/tree/main/lambdaKG, with a demo video at http://deepke.zjukg.cn/lambdakg.mp4 and long-term maintenance.
comment: Work in progress and the project website is https://zjunlp.github.io/project/promptkg/
♻ ☆ STACC: Code Comment Classification using SentenceTransformers
Code comments are a key resource for information about software artefacts. Depending on the use case, only some types of comments are useful. Thus, automatic approaches to classify these comments have been proposed. In this work, we address this need by proposing, STACC, a set of SentenceTransformers-based binary classifiers. These lightweight classifiers are trained and tested on the NLBSE Code Comment Classification tool competition dataset, and surpass the baseline by a significant margin, achieving an average F1 score of 0.74 against the baseline of 0.31, which is an improvement of 139%. A replication package, as well as the models themselves, are publicly available.
♻ ☆ ComSearch: Equation Searching with Combinatorial Strategy for Solving Math Word Problems with Weak Supervision EACL 2023
Previous studies have introduced a weakly-supervised paradigm for solving math word problems requiring only the answer value annotation. While these methods search for correct value equation candidates as pseudo labels, they search among a narrow sub-space of the enormous equation space. To address this problem, we propose a novel search algorithm with combinatorial strategy \textbf{ComSearch}, which can compress the search space by excluding mathematically equivalent equations. The compression allows the searching algorithm to enumerate all possible equations and obtain high-quality data. We investigate the noise in the pseudo labels that hold wrong mathematical logic, which we refer to as the \textit{false-matching} problem, and propose a ranking model to denoise the pseudo labels. Our approach holds a flexible framework to utilize two existing supervised math word problem solvers to train pseudo labels, and both achieve state-of-the-art performance in the weak supervision task.
comment: EACL 2023 long paper, 14 pages
♻ ☆ Perceive and predict: self-supervised speech representation based loss functions for speech enhancement ICASSP 2023
Recent work in the domain of speech enhancement has explored the use of self-supervised speech representations to aid in the training of neural speech enhancement models. However, much of this work focuses on using the deepest or final outputs of self supervised speech representation models, rather than the earlier feature encodings. The use of self supervised representations in such a way is often not fully motivated. In this work it is shown that the distance between the feature encodings of clean and noisy speech correlate strongly with psychoacoustically motivated measures of speech quality and intelligibility, as well as with human Mean Opinion Score (MOS) ratings. Experiments using this distance as a loss function are performed and improved performance over the use of STFT spectrogram distance based loss as well as other common loss functions from speech enhancement literature is demonstrated using objective measures such as perceptual evaluation of speech quality (PESQ) and short-time objective intelligibility (STOI).
comment: 4 pages, accepted at ICASSP 2023
♻ ☆ Knowledge-aware Bayesian Co-attention for Multimodal Emotion Recognition ICASSP 2023
Multimodal emotion recognition is a challenging research area that aims to fuse different modalities to predict human emotion. However, most existing models that are based on attention mechanisms have difficulty in learning emotionally relevant parts on their own. To solve this problem, we propose to incorporate external emotion-related knowledge in the co-attention based fusion of pre-trained models. To effectively incorporate this knowledge, we enhance the co-attention model with a Bayesian attention module (BAM) where a prior distribution is estimated using the emotion-related knowledge. Experimental results on the IEMOCAP dataset show that the proposed approach can outperform several state-of-the-art approaches by at least 0.7% unweighted accuracy (UA).
comment: Accepted to IEEE ICASSP 2023
♻ ☆ Can discrete information extraction prompts generalize across language models? ICLR 2023
We study whether automatically-induced prompts that effectively extract information from a language model can also be used, out-of-the-box, to probe other language models for the same information. After confirming that discrete prompts induced with the AutoPrompt algorithm outperform manual and semi-manual prompts on the slot-filling task, we demonstrate a drop in performance for AutoPrompt prompts learned on a model and tested on another. We introduce a way to induce prompts by mixing language models at training time that results in prompts that generalize well across models. We conduct an extensive analysis of the induced prompts, finding that the more general prompts include a larger proportion of existing English words and have a less order-dependent and more uniform distribution of information across their component tokens. Our work provides preliminary evidence that it's possible to generate discrete prompts that can be induced once and used with a number of different models, and gives insights on the properties characterizing such prompts.
comment: Published as conference paper at ICLR 2023
♻ ☆ Dialogue State Distillation Network with Inter-slot Contrastive Learning for Dialogue State Tracking AAAI 2023
In task-oriented dialogue systems, Dialogue State Tracking (DST) aims to extract users' intentions from the dialogue history. Currently, most existing approaches suffer from error propagation and are unable to dynamically select relevant information when utilizing previous dialogue states. Moreover, the relations between the updates of different slots provide vital clues for DST. However, the existing approaches rely only on predefined graphs to indirectly capture the relations. In this paper, we propose a Dialogue State Distillation Network (DSDN) to utilize relevant information of previous dialogue states and migrate the gap of utilization between training and testing. Thus, it can dynamically exploit previous dialogue states and avoid introducing error propagation simultaneously. Further, we propose an inter-slot contrastive learning loss to effectively capture the slot co-update relations from dialogue context. Experiments are conducted on the widely used MultiWOZ 2.0 and MultiWOZ 2.1 datasets. The experimental results show that our proposed model achieves the state-of-the-art performance for DST.
comment: Accepted by AAAI 2023
♻ ☆ Learning Deep Semantics for Test Completion
Writing tests is a time-consuming yet essential task during software development. We propose to leverage recent advances in deep learning for text and code generation to assist developers in writing tests. We formalize the novel task of test completion to automatically complete the next statement in a test method based on the context of prior statements and the code under test. We develop TeCo -- a deep learning model using code semantics for test completion. The key insight underlying TeCo is that predicting the next statement in a test method requires reasoning about code execution, which is hard to do with only syntax-level data that existing code completion models use. TeCo extracts and uses six kinds of code semantics data, including the execution result of prior statements and the execution context of the test method. To provide a testbed for this new task, as well as to evaluate TeCo, we collect a corpus of 130,934 test methods from 1,270 open-source Java projects. Our results show that TeCo achieves an exact-match accuracy of 18, which is 29% higher than the best baseline using syntax-level data only. When measuring functional correctness of generated next statement, TeCo can generate runnable code in 29% of the cases compared to 18% obtained by the best baseline. Moreover, TeCo is significantly better than prior work on test oracle generation.
comment: Accepted as a conference paper in ICSE 2023
♻ ☆ Can ChatGPT Assess Human Personalities? A General Evaluation Framework
Large Language Models (LLMs) especially ChatGPT have produced impressive results in various areas, but their potential human-like psychology is still largely unexplored. Existing works study the virtual personalities of LLMs but rarely explore the possibility of analyzing human personalities via LLMs. This paper presents a generic evaluation framework for LLMs to assess human personalities based on Myers Briggs Type Indicator (MBTI) tests. Specifically, we first devise unbiased prompts by randomly permuting options in MBTI questions and adopt the average testing result to encourage more impartial answer generation. Then, we propose to replace the subject in question statements to enable flexible queries and assessments on different subjects from LLMs. Finally, we re-formulate the question instructions in a manner of correctness evaluation to facilitate LLMs to generate clearer responses. The proposed framework enables LLMs to flexibly assess personalities of different groups of people. We further propose three evaluation metrics to measure the consistency, robustness, and fairness of assessment results from state-of-the-art LLMs including ChatGPT and InstructGPT. Our experiments reveal ChatGPT's ability to assess human personalities, and the average results demonstrate that it can achieve more consistent and fairer assessments in spite of lower robustness against prompt biases compared with InstructGPT.
comment: Our codes are available at https://github.com/Kali-Hac/ChatGPT-MBTI
♻ ☆ Effects of Parameter Norm Growth During Transformer Training: Inductive Bias from Gradient Descent EMNLP 2021
The capacity of neural networks like the widely adopted transformer is known to be very high. Evidence is emerging that they learn successfully due to inductive bias in the training routine, typically a variant of gradient descent (GD). To better understand this bias, we study the tendency for transformer parameters to grow in magnitude ($\ell_2$ norm) during training, and its implications for the emergent representations within self attention layers. Empirically, we document norm growth in the training of transformer language models, including T5 during its pretraining. As the parameters grow in magnitude, we prove that the network approximates a discretized network with saturated activation functions. Such "saturated" networks are known to have a reduced capacity compared to the full network family that can be described in terms of formal languages and automata. Our results suggest saturation is a new characterization of an inductive bias implicit in GD of particular interest for NLP. We leverage the emergent discrete structure in a saturated transformer to analyze the role of different attention heads, finding that some focus locally on a small number of positions, while other heads compute global averages, allowing counting. We believe understanding the interplay between these two capabilities may shed further light on the structure of computation within large transformers.
comment: Appeared at EMNLP 2021. March 7, 2023: Removed irreproducible numbers reported in a footnote with erratum note
♻ ☆ The Parallelism Tradeoff: Limitations of Log-Precision Transformers ACL
Despite their omnipresence in modern NLP, characterizing the computational power of transformer neural nets remains an interesting open question. We prove that transformers whose arithmetic precision is logarithmic in the number of input tokens (and whose feedforward nets are computable using space linear in their input) can be simulated by constant-depth logspace-uniform threshold circuits. This provides insight on the power of transformers using known results in complexity theory. For example, if $\mathsf L \neq \mathsf P$ (i.e., not all poly-time problems can be solved using logarithmic space), then transformers cannot even accurately solve linear equalities or check membership in an arbitrary context-free grammar with empty productions. Our result intuitively emerges from the transformer architecture's high parallelizability. We thus speculatively introduce the idea of a fundamental parallelism tradeoff: any model architecture as parallelizable as the transformer will obey limitations similar to it. Since parallelism is key to training models at massive scale, this suggests a potential inherent weakness of the scaling paradigm.
comment: Accepted at TACL. Formerly entitled "Log-Precision Transformers are Constant-Depth Threshold Circuits". Updated with minor corrections March 6, 2023
♻ ☆ Out-of-Distribution Detection and Selective Generation for Conditional Language Models ICLR 2023
Machine learning algorithms typically assume independent and identically distributed samples in training and at test time. Much work has shown that high-performing ML classifiers can degrade significantly and provide overly-confident, wrong classification predictions, particularly for out-of-distribution (OOD) inputs. Conditional language models (CLMs) are predominantly trained to classify the next token in an output sequence, and may suffer even worse degradation on OOD inputs as the prediction is done auto-regressively over many steps. Furthermore, the space of potential low-quality outputs is larger as arbitrary text can be generated and it is important to know when to trust the generated output. We present a highly accurate and lightweight OOD detection method for CLMs, and demonstrate its effectiveness on abstractive summarization and translation. We also show how our method can be used under the common and realistic setting of distribution shift for selective generation (analogous to selective prediction for classification) of high-quality outputs, while automatically abstaining from low-quality ones, enabling safer deployment of generative language models.
comment: Published in ICLR 2023
♻ ☆ AugTriever: Unsupervised Dense Retrieval by Scalable Data Augmentation
Dense retrievers have made significant strides in text retrieval and open-domain question answering, even though most achievements were made possible only with large amounts of human supervision. In this work, we aim to develop unsupervised methods by proposing two methods that create pseudo query-document pairs and train dense retrieval models in an annotation-free and scalable manner: query extraction and transferred query generation. The former method produces pseudo queries by selecting salient spans from the original document. The latter utilizes generation models trained for other NLP tasks (e.g., summarization) to produce pseudo queries. Extensive experiments show that models trained with the proposed augmentation methods can perform comparably well (or better) to multiple strong baselines. Combining those strategies leads to further improvements, achieving the state-of-the-art performance of unsupervised dense retrieval on both BEIR and ODQA datasets.
Computer Vision and Pattern Recognition
☆ Can We Scale Transformers to Predict Parameters of Diverse ImageNet Models?
Pretraining a neural network on a large dataset is becoming a cornerstone in machine learning that is within the reach of only a few communities with large-resources. We aim at an ambitious goal of democratizing pretraining. Towards that goal, we train and release a single neural network that can predict high quality ImageNet parameters of other neural networks. By using predicted parameters for initialization we are able to boost training of diverse ImageNet models available in PyTorch. When transferred to other datasets, models initialized with predicted parameters also converge faster and reach competitive final performance.
comment: Code and models are available at https://github.com/SamsungSAILMontreal/ghn3
☆ TrafficBots: Towards World Models for Autonomous Driving Simulation and Motion Prediction ICRA 2023
Data-driven simulation has become a favorable way to train and test autonomous driving algorithms. The idea of replacing the actual environment with a learned simulator has also been explored in model-based reinforcement learning in the context of world models. In this work, we show data-driven traffic simulation can be formulated as a world model. We present TrafficBots, a multi-agent policy built upon motion prediction and end-to-end driving, and based on TrafficBots we obtain a world model tailored for the planning module of autonomous vehicles. Existing data-driven traffic simulators are lacking configurability and scalability. To generate configurable behaviors, for each agent we introduce a destination as navigational information, and a time-invariant latent personality that specifies the behavioral style. To improve the scalability, we present a new scheme of positional encoding for angles, allowing all agents to share the same vectorized context and the use of an architecture based on dot-product attention. As a result, we can simulate all traffic participants seen in dense urban scenarios. Experiments on the Waymo open motion dataset show TrafficBots can simulate realistic multi-agent behaviors and achieve good performance on the motion prediction task.
comment: Accepted at ICRA 2023. The repository is available at https://github.com/SysCV/TrafficBots
☆ Predicted Embedding Power Regression for Large-Scale Out-of-Distribution Detection
Out-of-distribution (OOD) inputs can compromise the performance and safety of real world machine learning systems. While many methods exist for OOD detection and work well on small scale datasets with lower resolution and few classes, few methods have been developed for large-scale OOD detection. Existing large-scale methods generally depend on maximum classification probability, such as the state-of-the-art grouped softmax method. In this work, we develop a novel approach that calculates the probability of the predicted class label based on label distributions learned during the training process. Our method performs better than current state-of-the-art methods with only a negligible increase in compute cost. We evaluate our method against contemporary methods across $14$ datasets and achieve a statistically significant improvement with respect to AUROC (84.2 vs 82.4) and AUPR (96.2 vs 93.7).
☆ Introspective Cross-Attention Probing for Lightweight Transfer of Pre-trained Models
We propose InCA, a lightweight method for transfer learning that cross-attends to any activation layer of a pre-trained model. During training, InCA uses a single forward pass to extract multiple activations, which are passed to external cross-attention adapters, trained anew and combined or selected for downstream tasks. We show that, even when selecting a single top-scoring adapter, InCA achieves performance comparable to full fine-tuning, at a cost comparable to fine-tuning just the last layer. For example, with a cross-attention probe 1.3% the size of a pre-trained ViT-L/16 model, we achieve performance within 0.2% of the full fine-tuning paragon at 51% training cost of the baseline, on average across 11 downstream classification tasks. Unlike other forms of efficient adaptation, InCA does not require backpropagating through the pre-trained model, thus leaving its execution unaltered at both training and inference. The versatility of InCA is best illustrated in fine-grained tasks, which may require accessing information absent in the last layer but accessible in intermediate layer activations. Since the backbone is fixed, InCA allows parallel ensembling as well as parallel execution of multiple tasks. InCA achieves state-of-the-art performance in the ImageNet-to-Sketch multi-task benchmark.
☆ NEPHELE: A Neural Platform for Highly Realistic Cloud Radiance Rendering
We have recently seen tremendous progress in neural rendering (NR) advances, i.e., NeRF, for photo-real free-view synthesis. Yet, as a local technique based on a single computer/GPU, even the best-engineered Instant-NGP or i-NGP cannot reach real-time performance when rendering at a high resolution, and often requires huge local computing resources. In this paper, we resort to cloud rendering and present NEPHELE, a neural platform for highly realistic cloud radiance rendering. In stark contrast with existing NR approaches, our NEPHELE allows for more powerful rendering capabilities by combining multiple remote GPUs and facilitates collaboration by allowing multiple people to view the same NeRF scene simultaneously. We introduce i-NOLF to employ opacity light fields for ultra-fast neural radiance rendering in a one-query-per-ray manner. We further resemble the Lumigraph with geometry proxies for fast ray querying and subsequently employ a small MLP to model the local opacity lumishperes for high-quality rendering. We also adopt Perfect Spatial Hashing in i-NOLF to enhance cache coherence. As a result, our i-NOLF achieves an order of magnitude performance gain in terms of efficiency than i-NGP, especially for the multi-user multi-viewpoint setting under cloud rendering scenarios. We further tailor a task scheduler accompanied by our i-NOLF representation and demonstrate the advance of our methodological design through a comprehensive cloud platform, consisting of a series of cooperated modules, i.e., render farms, task assigner, frame composer, and detailed streaming strategies. Using such a cloud platform compatible with neural rendering, we further showcase the capabilities of our cloud radiance rendering through a series of applications, ranging from cloud VR/AR rendering.
☆ Meta-Explore: Exploratory Hierarchical Vision-and-Language Navigation Using Scene Object Spectrum Grounding CVPR 2023
The main challenge in vision-and-language navigation (VLN) is how to understand natural-language instructions in an unseen environment. The main limitation of conventional VLN algorithms is that if an action is mistaken, the agent fails to follow the instructions or explores unnecessary regions, leading the agent to an irrecoverable path. To tackle this problem, we propose Meta-Explore, a hierarchical navigation method deploying an exploitation policy to correct misled recent actions. We show that an exploitation policy, which moves the agent toward a well-chosen local goal among unvisited but observable states, outperforms a method which moves the agent to a previously visited state. We also highlight the demand for imagining regretful explorations with semantically meaningful clues. The key to our approach is understanding the object placements around the agent in spectral-domain. Specifically, we present a novel visual representation, called scene object spectrum (SOS), which performs category-wise 2D Fourier transform of detected objects. Combining exploitation policy and SOS features, the agent can correct its path by choosing a promising local goal. We evaluate our method in three VLN benchmarks: R2R, SOON, and REVERIE. Meta-Explore outperforms other baselines and shows significant generalization performance. In addition, local goal search using the proposed spectral-domain SOS features significantly improves the success rate by 17.1% and SPL by 20.6% for the SOON benchmark.
comment: Accepted by CVPR 2023. Project page: https://rllab-snu.github.io/projects/Meta-Explore/doc.html
☆ VOCALExplore: Pay-as-You-Go Video Data Exploration and Model Building
We introduce VOCALExplore, a system designed to support users in building domain-specific models over video datasets. VOCALExplore supports interactive labeling sessions and trains models using user-supplied labels. VOCALExplore maximizes model quality by automatically deciding how to select samples based on observed skew in the collected labels. It also selects the optimal video representations to use when training models by casting feature selection as a rising bandit problem. Finally, VOCALExplore implements optimizations to achieve low latency without sacrificing model performance. We demonstrate that VOCALExplore achieves close to the best possible model quality given candidate acquisition functions and feature extractors, and it does so with low visible latency (~1 second per iteration) and no expensive preprocessing.
☆ DeepSeeColor: Realtime Adaptive Color Correction for Autonomous Underwater Vehicles via Deep Learning Methods ICRA
Successful applications of complex vision-based behaviours underwater have lagged behind progress in terrestrial and aerial domains. This is largely due to the degraded image quality resulting from the physical phenomena involved in underwater image formation. Spectrally-selective light attenuation drains some colors from underwater images while backscattering adds others, making it challenging to perform vision-based tasks underwater. State-of-the-art methods for underwater color correction optimize the parameters of image formation models to restore the full spectrum of color to underwater imagery. However, these methods have high computational complexity that is unfavourable for realtime use by autonomous underwater vehicles (AUVs), as a result of having been primarily designed for offline color correction. Here, we present DeepSeeColor, a novel algorithm that combines a state-of-the-art underwater image formation model with the computational efficiency of deep learning frameworks. In our experiments, we show that DeepSeeColor offers comparable performance to the popular "Sea-Thru" algorithm (Akkaynak & Treibitz, 2019) while being able to rapidly process images at up to 60Hz, thus making it suitable for use onboard AUVs as a preprocessing step to enable more robust vision-based behaviours.
comment: 7 pages, 5 figures, 2 tables. Presented at the 2023 International Conference on Robotics and Automation (ICRA)
☆ Decoupling Skill Learning from Robotic Control for Generalizable Object Manipulation ICRA
Recent works in robotic manipulation through reinforcement learning (RL) or imitation learning (IL) have shown potential for tackling a range of tasks e.g., opening a drawer or a cupboard. However, these techniques generalize poorly to unseen objects. We conjecture that this is due to the high-dimensional action space for joint control. In this paper, we take an alternative approach and separate the task of learning 'what to do' from 'how to do it' i.e., whole-body control. We pose the RL problem as one of determining the skill dynamics for a disembodied virtual manipulator interacting with articulated objects. The whole-body robotic kinematic control is optimized to execute the high-dimensional joint motion to reach the goals in the workspace. It does so by solving a quadratic programming (QP) model with robotic singularity and kinematic constraints. Our experiments on manipulating complex articulated objects show that the proposed approach is more generalizable to unseen objects with large intra-class variations, outperforming previous approaches. The evaluation results indicate that our approach generates more compliant robotic motion and outperforms the pure RL and IL baselines in task success rates.
comment: Accepted to IEEE International Conference on Robotics and Automation (ICRA) 2023
☆ ELODIN: Naming Concepts in Embedding Spaces
Despite recent advancements, the field of text-to-image synthesis still suffers from lack of fine-grained control. Using only text, it remains challenging to deal with issues such as concept coherence and concept contamination. We propose a method to enhance control by generating specific concepts that can be reused throughout multiple images, effectively expanding natural language with new words that can be combined much like a painter's palette. Unlike previous contributions, our method does not copy visuals from input data and can generate concepts through text alone. We perform a set of comparisons that finds our method to be a significant improvement over text-only prompts.
☆ OpenOccupancy: A Large Scale Benchmark for Surrounding Semantic Occupancy Perception
Semantic occupancy perception is essential for autonomous driving, as automated vehicles require a fine-grained perception of the 3D urban structures. However, existing relevant benchmarks lack diversity in urban scenes, and they only evaluate front-view predictions. Towards a comprehensive benchmarking of surrounding perception algorithms, we propose OpenOccupancy, which is the first surrounding semantic occupancy perception benchmark. In the OpenOccupancy benchmark, we extend the large-scale nuScenes dataset with dense semantic occupancy annotations. Previous annotations rely on LiDAR points superimposition, where some occupancy labels are missed due to sparse LiDAR channels. To mitigate the problem, we introduce the Augmenting And Purifying (AAP) pipeline to ~2x densify the annotations, where ~4000 human hours are involved in the labeling process. Besides, camera-based, LiDAR-based and multi-modal baselines are established for the OpenOccupancy benchmark. Furthermore, considering the complexity of surrounding occupancy perception lies in the computational burden of high-resolution 3D predictions, we propose the Cascade Occupancy Network (CONet) to refine the coarse prediction, which relatively enhances the performance by ~30% than the baseline. We hope the OpenOccupancy benchmark will boost the development of surrounding occupancy perception algorithms.
comment: project page: https://github.com/JeffWang987/OpenOccupancy
☆ DINet: Deformation Inpainting Network for Realistic Face Visually Dubbing on High Resolution Video AAAI-23
For few-shot learning, it is still a critical challenge to realize photo-realistic face visually dubbing on high-resolution videos. Previous works fail to generate high-fidelity dubbing results. To address the above problem, this paper proposes a Deformation Inpainting Network (DINet) for high-resolution face visually dubbing. Different from previous works relying on multiple up-sample layers to directly generate pixels from latent embeddings, DINet performs spatial deformation on feature maps of reference images to better preserve high-frequency textural details. Specifically, DINet consists of one deformation part and one inpainting part. In the first part, five reference facial images adaptively perform spatial deformation to create deformed feature maps encoding mouth shapes at each frame, in order to align with the input driving audio and also the head poses of the input source images. In the second part, to produce face visually dubbing, a feature decoder is responsible for adaptively incorporating mouth movements from the deformed feature maps and other attributes (i.e., head pose and upper facial expression) from the source feature maps together. Finally, DINet achieves face visually dubbing with rich textural details. We conduct qualitative and quantitative comparisons to validate our DINet on high-resolution videos. The experimental results show that our method outperforms state-of-the-art works.
comment: AAAI-23, 9pages
☆ Comparing 3D deformations between longitudinal daily CBCT acquisitions using CNN for head and neck radiotherapy toxicity prediction
Adaptive radiotherapy is a growing field of study in cancer treatment due to it's objective in sparing healthy tissue. The standard of care in several institutions includes longitudinal cone-beam computed tomography (CBCT) acquisitions to monitor changes, but have yet to be used to improve tumor control while managing side-effects. The aim of this study is to demonstrate the clinical value of pre-treatment CBCT acquired daily during radiation therapy treatment for head and neck cancers for the downstream task of predicting severe toxicity occurrence: reactive feeding tube (NG), hospitalization and radionecrosis. For this, we propose a deformable 3D classification pipeline that includes a component analyzing the Jacobian matrix of the deformation between planning CT and longitudinal CBCT, as well as clinical data. The model is based on a multi-branch 3D residual convolutional neural network, while the CT to CBCT registration is based on a pair of VoxelMorph architectures. Accuracies of 85.8% and 75.3% was found for radionecrosis and hospitalization, respectively, with similar performance as early as after the first week of treatment. For NG tube risk, performance improves with increasing the timing of the CBCT fraction, reaching 83.1% after the $5_{th}$ week of treatment.
comment: 11 pages, 3 figures, 2 equations, 2 tables
☆ CUREE: A Curious Underwater Robot for Ecosystem Exploration
The current approach to exploring and monitoring complex underwater ecosystems, such as coral reefs, is to conduct surveys using diver-held or static cameras, or deploying sensor buoys. These approaches often fail to capture the full variation and complexity of interactions between different reef organisms and their habitat. The CUREE platform presented in this paper provides a unique set of capabilities in the form of robot behaviors and perception algorithms to enable scientists to explore different aspects of an ecosystem. Examples of these capabilities include low-altitude visual surveys, soundscape surveys, habitat characterization, and animal following. We demonstrate these capabilities by describing two field deployments on coral reefs in the US Virgin Islands. In the first deployment, we show that CUREE can identify the preferred habitat type of snapping shrimp in a reef through a combination of a visual survey, habitat characterization, and a soundscape survey. In the second deployment, we demonstrate CUREE's ability to follow arbitrary animals by separately following a barracuda and stingray for several minutes each in midwater and benthic environments, respectively.
comment: 7 pages
☆ FFT-based Dynamic Token Mixer for Vision
Multi-head-self-attention (MHSA)-equipped models have achieved notable performance in computer vision. Their computational complexity is proportional to quadratic numbers of pixels in input feature maps, resulting in slow processing, especially when dealing with high-resolution images. New types of token-mixer are proposed as an alternative to MHSA to circumvent this problem: an FFT-based token-mixer, similar to MHSA in global operation but with lower computational complexity. However, despite its attractive properties, the FFT-based token-mixer has not been carefully examined in terms of its compatibility with the rapidly evolving MetaFormer architecture. Here, we propose a novel token-mixer called dynamic filter and DFFormer and CDFFormer, image recognition models using dynamic filters to close the gaps above. CDFFormer achieved a Top-1 accuracy of 85.0%, close to the hybrid architecture with convolution and MHSA. Other wide-ranging experiments and analysis, including object detection and semantic segmentation, demonstrate that they are competitive with state-of-the-art architectures; Their throughput and memory efficiency when dealing with high-resolution image recognition is convolution and MHSA, not much different from ConvFormer, and far superior to CAFormer. Our results indicate that the dynamic filter is one of the token-mixer options that should be seriously considered. The code is available at https://github.com/okojoalg/dfformer
☆ A survey on automated detection and classification of acute leukemia and WBCs in microscopic blood cells
Leukemia (blood cancer) is an unusual spread of White Blood Cells or Leukocytes (WBCs) in the bone marrow and blood. Pathologists can diagnose leukemia by looking at a person's blood sample under a microscope. They identify and categorize leukemia by counting various blood cells and morphological features. This technique is time-consuming for the prediction of leukemia. The pathologist's professional skills and experiences may be affecting this procedure, too. In computer vision, traditional machine learning and deep learning techniques are practical roadmaps that increase the accuracy and speed in diagnosing and classifying medical images such as microscopic blood cells. This paper provides a comprehensive analysis of the detection and classification of acute leukemia and WBCs in the microscopic blood cells. First, we have divided the previous works into six categories based on the output of the models. Then, we describe various steps of detection and classification of acute leukemia and WBCs, including Data Augmentation, Preprocessing, Segmentation, Feature Extraction, Feature Selection (Reduction), Classification, and focus on classification step in the methods. Finally, we divide automated detection and classification of acute leukemia and WBCs into three categories, including traditional, Deep Neural Network (DNN), and mixture (traditional and DNN) methods based on the type of classifier in the classification step and analyze them. The results of this study show that in the diagnosis and classification of acute leukemia and WBCs, the Support Vector Machine (SVM) classifier in traditional machine learning models and Convolutional Neural Network (CNN) classifier in deep learning models have widely employed. The performance metrics of the models that use these classifiers compared to the others model are higher.
☆ InsMOS: Instance-Aware Moving Object Segmentation in LiDAR Data
Identifying moving objects is a crucial capability for autonomous navigation, consistent map generation, and future trajectory prediction of objects. In this paper, we propose a novel network that addresses the challenge of segmenting moving objects in 3D LiDAR scans. Our approach not only predicts point-wise moving labels but also detects instance information of main traffic participants. Such a design helps determine which instances are actually moving and which ones are temporarily static in the current scene. Our method exploits a sequence of point clouds as input and quantifies them into 4D voxels. We use 4D sparse convolutions to extract motion features from the 4D voxels and inject them into the current scan. Then, we extract spatio-temporal features from the current scan for instance detection and feature fusion. Finally, we design an upsample fusion module to output point-wise labels by fusing the spatio-temporal features and predicted instance information. We evaluated our approach on the LiDAR-MOS benchmark based on SemanticKITTI and achieved better moving object segmentation performance compared to state-of-the-art methods, demonstrating the effectiveness of our approach in integrating instance information for moving object segmentation. Furthermore, our method shows superior performance on the Apollo dataset with a pre-trained model on SemanticKITTI, indicating that our method generalizes well in different scenes.The code and pre-trained models of our method will be released at https://github.com/nubot-nudt/InsMOS.
comment: 8 pages, 4 figures, submitted
☆ SpinDOE: A ball spin estimation method for table tennis robot
Spin plays a considerable role in table tennis, making a shot's trajectory harder to read and predict. However, the spin is challenging to measure because of the ball's high velocity and the magnitude of the spin values. Existing methods either require extremely high framerate cameras or are unreliable because they use the ball's logo, which may not always be visible. Because of this, many table tennis-playing robots ignore the spin, which severely limits their capabilities. This paper proposes an easily implementable and reliable spin estimation method. We developed a dotted-ball orientation estimation (DOE) method, that can then be used to estimate the spin. The dots are first localized on the image using a CNN and then identified using geometric hashing. The spin is finally regressed from the estimated orientations. Using our algorithm, the ball's orientation can be estimated with a mean error of 2.4{\deg} and the spin estimation has an relative error lower than 1%. Spins up to 175 rps are measurable with a camera of 350 fps in real time. Using our method, we generated a dataset of table tennis ball trajectories with position and spin, available on our project page.
☆ Organelle-specific segmentation, spatial analysis, and visualization of volume electron microscopy datasets
Volume electron microscopy is the method of choice for the in-situ interrogation of cellular ultrastructure at the nanometer scale. Recent technical advances have led to a rapid increase in large raw image datasets that require computational strategies for segmentation and spatial analysis. In this protocol, we describe a practical and annotation-efficient pipeline for organelle-specific segmentation, spatial analysis, and visualization of large volume electron microscopy datasets using freely available, user-friendly software tools that can be run on a single standard workstation. We specifically target researchers in the life sciences with limited computational expertise, who face the following tasks within their volume electron microscopy projects: i) How to generate 3D segmentation labels for different types of cell organelles while minimizing manual annotation efforts, ii) how to analyze the spatial interactions between organelle instances, and iii) how to best visualize the 3D segmentation results. To meet these demands we give detailed guidelines for choosing the most efficient segmentation tools for the specific cell organelle. We furthermore provide easily executable components for spatial analysis and 3D rendering and bridge compatibility issues between freely available open-source tools, such that others can replicate our full pipeline starting from a raw dataset up to the final plots and rendered images. We believe that our detailed description can serve as a valuable reference for similar projects requiring special strategies for single- or multiple organelle analysis which can be achieved with computational resources commonly available to single-user setups.
☆ Event Voxel Set Transformer for Spatiotemporal Representation Learning on Event Streams
Event cameras are neuromorphic vision sensors representing visual information as sparse and asynchronous event streams. Most state-of-the-art event-based methods project events into dense frames and process them with conventional learning models. However, these approaches sacrifice the sparsity and high temporal resolution of event data, resulting in a large model size and high computational complexity. To fit the sparse nature of events and sufficiently explore their implicit relationship, we develop a novel attention-aware framework named Event Voxel Set Transformer (EVSTr) for spatiotemporal representation learning on event streams. It first converts the event stream into a voxel set and then hierarchically aggregates voxel features to obtain robust representations. The core of EVSTr is an event voxel transformer encoder to extract discriminative spatiotemporal features, which consists of two well-designed components, including a multi-scale neighbor embedding layer (MNEL) for local information aggregation and a voxel self-attention layer (VSAL) for global representation modeling. Enabling the framework to incorporate a long-term temporal structure, we introduce a segmental consensus strategy for modeling motion patterns over a sequence of segmented voxel sets. We evaluate the proposed framework on two event-based tasks: object classification and action recognition. Comprehensive experiments show that EVSTr achieves state-of-the-art performance while maintaining low model complexity. Additionally, we present a new dataset (NeuroHAR) recorded in challenging visual scenarios to address the lack of real-world event-based datasets for action recognition.
comment: 12 pages, 7 figures
☆ Parsing Line Segments of Floor Plan Images Using Graph Neural Networks
In this paper, we present a GNN-based Line Segment Parser (GLSP), which uses a junction heatmap to predict line segments' endpoints, and graph neural networks to extract line segments and their categories. Different from previous floor plan recognition methods, which rely on semantic segmentation, our proposed method is able to output vectorized line segment and requires less post-processing steps to be put into practical use. Our experiments show that the methods outperform state-of-the-art line segment detection models on multi-class line segment detection tasks with floor plan images. In the paper, we use our floor plan dataset named Large-scale Residential Floor Plan data (LRFP). The dataset contains a total of 271,035 floor plan images. The label corresponding to each picture contains the scale information, the categories and outlines of rooms, and the endpoint positions of line segments such as doors, windows, and walls. Our augmentation method makes the dataset adaptable to the drawing styles of as many countries and regions as possible.
☆ Region and Spatial Aware Anomaly Detection for Fundus Images
Recently anomaly detection has drawn much attention in diagnosing ocular diseases. Most existing anomaly detection research in fundus images has relatively large anomaly scores in the salient retinal structures, such as blood vessels, optical cups and discs. In this paper, we propose a Region and Spatial Aware Anomaly Detection (ReSAD) method for fundus images, which obtains local region and long-range spatial information to reduce the false positives in the normal structure. ReSAD transfers a pre-trained model to extract the features of normal fundus images and applies the Region-and-Spatial-Aware feature Combination module (ReSC) for pixel-level features to build a memory bank. In the testing phase, ReSAD uses the memory bank to determine out-of-distribution samples as abnormalities. Our method significantly outperforms the existing anomaly detection methods for fundus images on two publicly benchmark datasets.
☆ Multiscale Tensor Decomposition and Rendering Equation Encoding for View Synthesis
Rendering novel views from captured multi-view images has made considerable progress since the emergence of the neural radiance field. This paper aims to further advance the quality of view rendering by proposing a novel approach dubbed the neural radiance feature field (NRFF) which represents scenes in the feature space. We first propose a multiscale tensor decomposition scheme to organize learnable features so as to represent scenes from coarse to fine scales. We demonstrate many benefits of the proposed multiscale representation, including more accurate scene shape and appearance reconstruction, and faster convergence compared with the single-scale representation. Instead of encoding view directions to model view-dependent effects, we further propose to encode the rendering equation in the feature space by employing the anisotropic spherical Gaussian mixture predicted from the proposed multiscale representation. The proposed NRFF improves state-of-the-art rendering results by over 1 dB in PSNR on both the NeRF and NSVF synthetic datasets. A significant improvement has also been observed on the real-world Tanks and Temples dataset.
☆ Lformer: Text-to-Image Generation with L-shape Block Parallel Decoding
Generative transformers have shown their superiority in synthesizing high-fidelity and high-resolution images, such as good diversity and training stability. However, they suffer from the problem of slow generation since they need to generate a long token sequence autoregressively. To better accelerate the generative transformers while keeping good generation quality, we propose Lformer, a semi-autoregressive text-to-image generation model. Lformer firstly encodes an image into $h{\times}h$ discrete tokens, then divides these tokens into $h$ mirrored L-shape blocks from the top left to the bottom right and decodes the tokens in a block parallelly in each step. Lformer predicts the area adjacent to the previous context like autoregressive models thus it is more stable while accelerating. By leveraging the 2D structure of image tokens, Lformer achieves faster speed than the existing transformer-based methods while keeping good generation quality. Moreover, the pretrained Lformer can edit images without the requirement for finetuning. We can roll back to the early steps for regeneration or edit the image with a bounding box and a text prompt.
☆ External Camera-based Mobile Robot Pose Estimation for Collaborative Perception with Smart Edge Sensors ICRA 2023
We present an approach for estimating a mobile robot's pose w.r.t. the allocentric coordinates of a network of static cameras using multi-view RGB images. The images are processed online, locally on smart edge sensors by deep neural networks to detect the robot and estimate 2D keypoints defined at distinctive positions of the 3D robot model. Robot keypoint detections are synchronized and fused on a central backend, where the robot's pose is estimated via multi-view minimization of reprojection errors. Through the pose estimation from external cameras, the robot's localization can be initialized in an allocentric map from a completely unknown state (kidnapped robot problem) and robustly tracked over time. We conduct a series of experiments evaluating the accuracy and robustness of the camera-based pose estimation compared to the robot's internal navigation stack, showing that our camera-based method achieves pose errors below 3 cm and 1{\deg} and does not drift over time, as the robot is localized allocentrically. With the robot's pose precisely estimated, its observations can be fused into the allocentric scene model. We show a real-world application, where observations from mobile robot and static smart edge sensors are fused to collaboratively build a 3D semantic map of a $\sim$240 m$^2$ indoor environment.
comment: Accepted for ICRA 2023, 7 pages, 8 figures
☆ Hidden Knowledge: Mathematical Methods for the Extraction of the Fingerprint of Medieval Paper from Digital Images
Medieval paper, a handmade product, is made with a mould which leaves an indelible imprint on the sheet of paper. This imprint includes chain lines, laid lines and watermarks which are often visible on the sheet. Extracting these features allows the identification of paper stock and gives information about chronology, localisation and movement of books and people. Most computational work for feature extraction of paper analysis has so far focused on radiography or transmitted light images. While these imaging methods provide clear visualisation for the features of interest, they are expensive and time consuming in their acquisition and not feasible for smaller institutions. However, reflected light images of medieval paper manuscripts are abundant and possibly cheaper in their acquisition. In this paper, we propose algorithms to detect and extract the laid and chain lines from reflected light images. We tackle the main drawback of reflected light images, that is, the low contrast attenuation of lines and intensity jumps due to noise and degradation, by employing the spectral total variation decomposition and develop methods for subsequent line extraction. Our results clearly demonstrate the feasibility of using reflected light images in paper analysis. This work enables the feature extraction for paper manuscripts that have otherwise not been analysed due to a lack of appropriate images. We also open the door for paper stock identification at scale.
☆ Guiding Pseudo-labels with Uncertainty Estimation for Test-Time Adaptation CVPR
Standard Unsupervised Domain Adaptation (UDA) methods assume the availability of both source and target data during the adaptation. In this work, we investigate the Test-Time Adaptation (TTA), a specific case of UDA where a model is adapted to a target domain without access to source data. We propose a novel approach for the TTA setting based on a loss reweighting strategy that brings robustness against the noise that inevitably affects the pseudo-labels. The classification loss is reweighted based on the reliability of the pseudo-labels that is measured by estimating their uncertainty. Guided by such reweighting strategy, the pseudo-labels are progressively refined by aggregating knowledge from neighbouring samples. Furthermore, a self-supervised contrastive framework is leveraged as a target space regulariser to enhance such knowledge aggregation. A novel negative pairs exclusion strategy is proposed to identify and exclude negative pairs made of samples sharing the same class, even in presence of some noise in the pseudo-labels. Our method outperforms previous methods on three major benchmarks by a large margin. We set the new TTA state-of-the-art on VisDA-C and DomainNet with a performance gain of +1.8\% on both benchmarks and on PACS with +12.3\% in the single-source setting and +6.6\% in\ multi-target adaptation. Additional analyses demonstrate that the proposed approach is robust to the noise, which results in significantly more accurate pseudo-labels compared to state-of-the-art approaches.
comment: To be published in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2023
☆ Proactive Multi-Camera Collaboration For 3D Human Pose Estimation ICLR 2023
This paper presents a multi-agent reinforcement learning (MARL) scheme for proactive Multi-Camera Collaboration in 3D Human Pose Estimation in dynamic human crowds. Traditional fixed-viewpoint multi-camera solutions for human motion capture (MoCap) are limited in capture space and susceptible to dynamic occlusions. Active camera approaches proactively control camera poses to find optimal viewpoints for 3D reconstruction. However, current methods still face challenges with credit assignment and environment dynamics. To address these issues, our proposed method introduces a novel Collaborative Triangulation Contribution Reward (CTCR) that improves convergence and alleviates multi-agent credit assignment issues resulting from using 3D reconstruction accuracy as the shared reward. Additionally, we jointly train our model with multiple world dynamics learning tasks to better capture environment dynamics and encourage anticipatory behaviors for occlusion avoidance. We evaluate our proposed method in four photo-realistic UE4 environments to ensure validity and generalizability. Empirical results show that our method outperforms fixed and active baselines in various scenarios with different numbers of cameras and humans.
comment: ICLR 2023 poster
☆ Graph Neural Networks in Vision-Language Image Understanding: A Survey
2D image understanding is a complex problem within Computer Vision, but it holds the key to providing human level scene comprehension. It goes further than identifying the objects in an image, and instead it attempts to understand the scene. Solutions to this problem form the underpinning of a range of tasks, including image captioning, Visual Question Answering (VQA), and image retrieval. Graphs provide a natural way to represent the relational arrangement between objects in an image, and thus in recent years Graph Neural Networks (GNNs) have become a standard component of many 2D image understanding pipelines, becoming a core architectural component especially in the VQA group of tasks. In this survey, we review this rapidly evolving field and we provide a taxonomy of graph types used in 2D image understanding approaches, a comprehensive list of the GNN models used in this domain, and a roadmap of future potential developments. To the best of our knowledge, this is the first comprehensive survey that covers image captioning, visual question answering, and image retrieval techniques that focus on using GNNs as the main part of their architecture.
comment: 19 pages, 5 figures, 6 tables
☆ Patched Diffusion Models for Unsupervised Anomaly Detection in Brain MRI
The use of supervised deep learning techniques to detect pathologies in brain MRI scans can be challenging due to the diversity of brain anatomy and the need for annotated data sets. An alternative approach is to use unsupervised anomaly detection, which only requires sample-level labels of healthy brains to create a reference representation. This reference representation can then be compared to unhealthy brain anatomy in a pixel-wise manner to identify abnormalities. To accomplish this, generative models are needed to create anatomically consistent MRI scans of healthy brains. While recent diffusion models have shown promise in this task, accurately generating the complex structure of the human brain remains a challenge. In this paper, we propose a method that reformulates the generation task of diffusion models as a patch-based estimation of healthy brain anatomy, using spatial context to guide and improve reconstruction. We evaluate our approach on data of tumors and multiple sclerosis lesions and demonstrate a relative improvement of 25.1% compared to existing baselines.
comment: Accepted full paper at the MIDL23 conference
☆ Deep Learning for Inertial Positioning: A Survey
Inertial sensor has been widely deployed on smartphones, drones, robots and IoT devices. Due to its importance in ubiquitous and robust localization, inertial sensor based positioning is key in many applications, including personal navigation, location based security, and human-device interaction. However, inertial positioning suffers from the so-called error drifts problem, as the measurements of low-cost MEMS inertial sensor are corrupted with various inevitable error sources, leading to unbounded drifts when being integrated doubly in traditional inertial navigation algorithms. Recently, with increasing sensor data and computational power, the fast developments in deep learning have spurred a large amount of research works in introducing deep learning to tackle the problem of inertial positioning. Relevant literature spans from the areas of mobile computing, robotics and machine learning. This article comprehensively reviews relevant works on deep learning based inertial positioning, connects the efforts from different fields, and covers how deep learning can be applied to solve sensor calibration, positioning error drifts reduction and sensor fusion. Finally, we provide insights on the benefits and limitations of existing works, and indicate the future opportunities in this direction.
☆ DLT: Conditioned layout generation with Joint Discrete-Continuous Diffusion Layout Transformer
Generating visual layouts is an essential ingredient of graphic design. The ability to condition layout generation on a partial subset of component attributes is critical to real-world applications that involve user interaction. Recently, diffusion models have demonstrated high-quality generative performances in various domains. However, it is unclear how to apply diffusion models to the natural representation of layouts which consists of a mix of discrete (class) and continuous (location, size) attributes. To address the conditioning layout generation problem, we introduce DLT, a joint discrete-continuous diffusion model. DLT is a transformer-based model which has a flexible conditioning mechanism that allows for conditioning on any given subset of all the layout component classes, locations, and sizes. Our method outperforms state-of-the-art generative models on various layout generation datasets with respect to different metrics and conditioning settings. Additionally, we validate the effectiveness of our proposed conditioning mechanism and the joint continuous-diffusion process. This joint process can be incorporated into a wide range of mixed discrete-continuous generative tasks.
☆ At Your Fingertips: Extracting Piano Fingering Instructions from Videos
Piano fingering -- knowing which finger to use to play each note in a musical piece, is a hard and important skill to master when learning to play the piano. While some sheet music is available with expert-annotated fingering information, most pieces lack this information, and people often resort to learning the fingering from demonstrations in online videos. We consider the AI task of automating the extraction of fingering information from videos. This is a non-trivial task as fingers are often occluded by other fingers, and it is often not clear from the video which of the keys were pressed, requiring the synchronization of hand position information and knowledge about the notes that were played. We show how to perform this task with high-accuracy using a combination of deep-learning modules, including a GAN-based approach for fine-tuning on out-of-domain data. We extract the fingering information with an f1 score of 97\%. We run the resulting system on 90 videos, resulting in high-quality piano fingering information of 150K notes, the largest available dataset of piano-fingering to date.
comment: 6 pages, paper from 2019
☆ LORE: Logical Location Regression Network for Table Structure Recognition
Table structure recognition (TSR) aims at extracting tables in images into machine-understandable formats. Recent methods solve this problem by predicting the adjacency relations of detected cell boxes, or learning to generate the corresponding markup sequences from the table images. However, they either count on additional heuristic rules to recover the table structures, or require a huge amount of training data and time-consuming sequential decoders. In this paper, we propose an alternative paradigm. We model TSR as a logical location regression problem and propose a new TSR framework called LORE, standing for LOgical location REgression network, which for the first time combines logical location regression together with spatial location regression of table cells. Our proposed LORE is conceptually simpler, easier to train and more accurate than previous TSR models of other paradigms. Experiments on standard benchmarks demonstrate that LORE consistently outperforms prior arts. Code is available at https:// github.com/AlibabaResearch/AdvancedLiterateMachinery/tree/main/DocumentUnderstanding/LORE-TSR.
☆ Learning Discriminative Representations for Skeleton Based Action Recognition CVPR2023
Human action recognition aims at classifying the category of human action from a segment of a video. Recently, people dive into designing GCN-based models to extract features from skeletons for performing this task, because skeleton representations are much efficient and robust than other modalities such as RGB frames. However, when employing the skeleton data, some important clues like related items are also dismissed. It results in some ambiguous actions that are hard to be distinguished and tend to be misclassified. To alleviate this problem, we propose an auxiliary feature refinement head (FR Head), which consists of spatial-temporal decoupling and contrastive feature refinement, to obtain discriminative representations of skeletons. Ambiguous samples are dynamically discovered and calibrated in the feature space. Furthermore, FR Head could be imposed on different stages of GCNs to build a multi-level refinement for stronger supervision. Extensive experiments are conducted on NTU RGB+D, NTU RGB+D 120, and NW-UCLA datasets. Our proposed models obtain competitive results from state-of-the-art methods and can help to discriminate those ambiguous samples.
comment: Accepted by CVPR2023. 10 pages, 5 figures, 5 tables
☆ Refined Pseudo labeling for Source-free Domain Adaptive Object Detection ICASSP 2023
Domain adaptive object detection (DAOD) assumes that both labeled source data and unlabeled target data are available for training, but this assumption does not always hold in real-world scenarios. Thus, source-free DAOD is proposed to adapt the source-trained detectors to target domains with only unlabeled target data. Existing source-free DAOD methods typically utilize pseudo labeling, where the performance heavily relies on the selection of confidence threshold. However, most prior works adopt a single fixed threshold for all classes to generate pseudo labels, which ignore the imbalanced class distribution, resulting in biased pseudo labels. In this work, we propose a refined pseudo labeling framework for source-free DAOD. First, to generate unbiased pseudo labels, we present a category-aware adaptive threshold estimation module, which adaptively provides the appropriate threshold for each category. Second, to alleviate incorrect box regression, a localization-aware pseudo label assignment strategy is introduced to divide labels into certain and uncertain ones and optimize them separately. Finally, extensive experiments on four adaptation tasks demonstrate the effectiveness of our method.
comment: Accepted to ICASSP 2023
☆ Challenges of the Creation of a Dataset for Vision Based Human Hand Action Recognition in Industrial Assembly
This work presents the Industrial Hand Action Dataset V1, an industrial assembly dataset consisting of 12 classes with 459,180 images in the basic version and 2,295,900 images after spatial augmentation. Compared to other freely available datasets tested, it has an above-average duration and, in addition, meets the technical and legal requirements for industrial assembly lines. Furthermore, the dataset contains occlusions, hand-object interaction, and various fine-grained human hand actions for industrial assembly tasks that were not found in combination in examined datasets. The recorded ground truth assembly classes were selected after extensive observation of real-world use cases. A Gated Transformer Network, a state-of-the-art model from the transformer domain was adapted, and proved with a test accuracy of 86.25% before hyperparameter tuning by 18,269,959 trainable parameters, that it is possible to train sequential deep learning models with this dataset.
☆ Bootstrap The Original Latent: Freeze-and-thaw Adapter for Back-Propagated Black-Box Adaptation
In this paper, considering the balance of data/model privacy of model owners and user needs, we propose a new setting called Back-Propagated Black-Box Adaptation (BPBA) for users to better train their private models via the guidance of the back-propagated results of foundation/source models. Our setting can ease the usage of foundation/source models as well as prevent the leakage and misuse of foundation/source models. Moreover, we also propose a new training strategy called Bootstrap The Original Latent (BTOL) to fully utilize the foundation/source models. Our strategy consists of a domain adapter and a freeze-and-thaw strategy. We apply our BTOL under BPBA and Black-box UDA settings on three different datasets. Experiments show that our strategy is efficient and robust in various settings without manual augmentations.
☆ Hybrid quantum-classical convolutional neural network for phytoplankton classification
The taxonomic composition and abundance of phytoplankton, having direct impact on marine ecosystem dynamic and global environment change, are listed as essential ocean variables. Phytoplankton classification is very crucial for Phytoplankton analysis, but it is very difficult because of the huge amount and tiny volume of Phytoplankton. Machine learning is the principle way of performing phytoplankton image classification automatically. When carrying out large-scale research on the marine phytoplankton, the volume of data increases overwhelmingly and more powerful computational resources are required for the success of machine learning algorithms. Recently, quantum machine learning has emerged as the potential solution for large-scale data processing by harnessing the exponentially computational power of quantum computer. Here, for the first time, we demonstrate the feasibility of quantum deep neural networks for phytoplankton classification. Hybrid quantum-classical convolutional and residual neural networks are developed based on the classical architectures. These models make a proper balance between the limited function of the current quantum devices and the large size of phytoplankton images, which make it possible to perform phytoplankton classification on the near-term quantum computers. Better performance is obtained by the quantum-enhanced models against the classical counterparts. In particular, quantum models converge much faster than classical ones. The present quantum models are versatile, and can be applied for various tasks of image classification in the field of marine science.
comment: 20 pages, 13 figures
☆ FIT: Frequency-based Image Translation for Domain Adaptive Object Detection ICONIP 2022
Domain adaptive object detection (DAOD) aims to adapt the detector from a labelled source domain to an unlabelled target domain. In recent years, DAOD has attracted massive attention since it can alleviate performance degradation due to the large shift of data distributions in the wild. To align distributions between domains, adversarial learning is widely used in existing DAOD methods. However, the decision boundary for the adversarial domain discriminator may be inaccurate, causing the model biased towards the source domain. To alleviate this bias, we propose a novel Frequency-based Image Translation (FIT) framework for DAOD. First, by keeping domain-invariant frequency components and swapping domain-specific ones, we conduct image translation to reduce domain shift at the input level. Second, hierarchical adversarial feature learning is utilized to further mitigate the domain gap at the feature level. Finally, we design a joint loss to train the entire network in an end-to-end manner without extra training to obtain translated images. Extensive experiments on three challenging DAOD benchmarks demonstrate the effectiveness of our method.
comment: Accepted to ICONIP 2022
☆ MOSO: Decomposing MOtion, Scene and Object for Video Prediction CVPR 2023
Motion, scene and object are three primary visual components of a video. In particular, objects represent the foreground, scenes represent the background, and motion traces their dynamics. Based on this insight, we propose a two-stage MOtion, Scene and Object decomposition framework (MOSO) for video prediction, consisting of MOSO-VQVAE and MOSO-Transformer. In the first stage, MOSO-VQVAE decomposes a previous video clip into the motion, scene and object components, and represents them as distinct groups of discrete tokens. Then, in the second stage, MOSO-Transformer predicts the object and scene tokens of the subsequent video clip based on the previous tokens and adds dynamic motion at the token level to the generated object and scene tokens. Our framework can be easily extended to unconditional video generation and video frame interpolation tasks. Experimental results demonstrate that our method achieves new state-of-the-art performance on five challenging benchmarks for video prediction and unconditional video generation: BAIR, RoboNet, KTH, KITTI and UCF101. In addition, MOSO can produce realistic videos by combining objects and scenes from different videos.
comment: Accepted by CVPR 2023
☆ Logit Margin Matters: Improving Transferable Targeted Adversarial Attack by Logit Calibration
Previous works have extensively studied the transferability of adversarial samples in untargeted black-box scenarios. However, it still remains challenging to craft targeted adversarial examples with higher transferability than non-targeted ones. Recent studies reveal that the traditional Cross-Entropy (CE) loss function is insufficient to learn transferable targeted adversarial examples due to the issue of vanishing gradient. In this work, we provide a comprehensive investigation of the CE loss function and find that the logit margin between the targeted and untargeted classes will quickly obtain saturation in CE, which largely limits the transferability. Therefore, in this paper, we devote to the goal of continually increasing the logit margin along the optimization to deal with the saturation issue and propose two simple and effective logit calibration methods, which are achieved by downscaling the logits with a temperature factor and an adaptive margin, respectively. Both of them can effectively encourage optimization to produce a larger logit margin and lead to higher transferability. Besides, we show that minimizing the cosine distance between the adversarial examples and the classifier weights of the target class can further improve the transferability, which is benefited from downscaling logits via L2-normalization. Experiments conducted on the ImageNet dataset validate the effectiveness of the proposed methods, which outperform the state-of-the-art methods in black-box targeted attacks. The source code is available at \href{https://github.com/WJJLL/Target-Attack/}{Link}
☆ MAST: Masked Augmentation Subspace Training for Generalizable Self-Supervised Priors ICLR 2023
Recent Self-Supervised Learning (SSL) methods are able to learn feature representations that are invariant to different data augmentations, which can then be transferred to downstream tasks of interest. However, different downstream tasks require different invariances for their best performance, so the optimal choice of augmentations for SSL depends on the target task. In this paper, we aim to learn self-supervised features that generalize well across a variety of downstream tasks (e.g., object classification, detection and instance segmentation) without knowing any task information beforehand. We do so by Masked Augmentation Subspace Training (or MAST) to encode in the single feature space the priors from different data augmentations in a factorized way. Specifically, we disentangle the feature space into separate subspaces, each induced by a learnable mask that selects relevant feature dimensions to model invariance to a specific augmentation. We show the success of MAST in jointly capturing generalizable priors from different augmentations, using both unique and shared features across the subspaces. We further show that MAST benefits from uncertainty modeling to reweight ambiguous samples from strong augmentations that may cause similarity mismatch in each subspace. Experiments demonstrate that MAST consistently improves generalization on various downstream tasks, while being task-agnostic and efficient during SSL. We also provide interesting insights about how different augmentations are related and how uncertainty reflects learning difficulty.
comment: ICLR 2023
☆ CIFF-Net: Contextual Image Feature Fusion for Melanoma Diagnosis
Melanoma is considered to be the deadliest variant of skin cancer causing around 75\% of total skin cancer deaths. To diagnose Melanoma, clinicians assess and compare multiple skin lesions of the same patient concurrently to gather contextual information regarding the patterns, and abnormality of the skin. So far this concurrent multi-image comparative method has not been explored by existing deep learning-based schemes. In this paper, based on contextual image feature fusion (CIFF), a deep neural network (CIFF-Net) is proposed, which integrates patient-level contextual information into the traditional approaches for improved Melanoma diagnosis by concurrent multi-image comparative method. The proposed multi-kernel self attention (MKSA) module offers better generalization of the extracted features by introducing multi-kernel operations in the self attention mechanisms. To utilize both self attention and contextual feature-wise attention, an attention guided module named contextual feature fusion (CFF) is proposed that integrates extracted features from different contextual images into a single feature vector. Finally, in comparative contextual feature fusion (CCFF) module, primary and contextual features are compared concurrently to generate comparative features. Significant improvement in performance has been achieved on the ISIC-2020 dataset over the traditional approaches that validate the effectiveness of the proposed contextual learning scheme.
☆ Run, Don't Walk: Chasing Higher FLOPS for Faster Neural Networks CVPR 2023
To design fast neural networks, many works have been focusing on reducing the number of floating-point operations (FLOPs). We observe that such reduction in FLOPs, however, does not necessarily lead to a similar level of reduction in latency. This mainly stems from inefficiently low floating-point operations per second (FLOPS). To achieve faster networks, we revisit popular operators and demonstrate that such low FLOPS is mainly due to frequent memory access of the operators, especially the depthwise convolution. We hence propose a novel partial convolution (PConv) that extracts spatial features more efficiently, by cutting down redundant computation and memory access simultaneously. Building upon our PConv, we further propose FasterNet, a new family of neural networks, which attains substantially higher running speed than others on a wide range of devices, without compromising on accuracy for various vision tasks. For example, on ImageNet-1k, our tiny FasterNet-T0 is $3.1\times$, $3.1\times$, and $2.5\times$ faster than MobileViT-XXS on GPU, CPU, and ARM processors, respectively, while being $2.9\%$ more accurate. Our large FasterNet-L achieves impressive $83.5\%$ top-1 accuracy, on par with the emerging Swin-B, while having $49\%$ higher inference throughput on GPU, as well as saving $42\%$ compute time on CPU. Code is available at \url{https://github.com/JierunChen/FasterNet}.
comment: Accepted to CVPR 2023
☆ F2BEV: Bird's Eye View Generation from Surround-View Fisheye Camera Images for Automated Driving
Bird's Eye View (BEV) representations are tremendously useful for perception-related automated driving tasks. However, generating BEVs from surround-view fisheye camera images is challenging due to the strong distortions introduced by such wide-angle lenses. We take the first step in addressing this challenge and introduce a baseline, F2BEV, to generate BEV height maps and semantic segmentation maps from fisheye images. F2BEV consists of a distortion-aware spatial cross attention module for querying and consolidating spatial information from fisheye image features in a transformer-style architecture followed by a task-specific head. We evaluate single-task and multi-task variants of F2BEV on our synthetic FB-SSEM dataset, all of which generate better BEV height and segmentation maps (in terms of the IoU) than a state-of-the-art BEV generation method operating on undistorted fisheye images. We also demonstrate height map generation from real-world fisheye images using F2BEV. An initial sample of our dataset is publicly available at https://tinyurl.com/58jvnscy
☆ Filter Pruning based on Information Capacity and Independence
Filter pruning has been widely used in the compression and acceleration of convolutional neural networks (CNNs). However, most existing methods are still challenged by heavy compute cost and biased filter selection. Moreover, most designs for filter evaluation miss interpretability due to the lack of appropriate theoretical guidance. In this paper, we propose a novel filter pruning method which evaluates filters in a interpretable, multi-persepective and data-free manner. We introduce information capacity, a metric that represents the amount of information contained in a filter. Based on the interpretability and validity of information entropy, we propose to use that as a quantitative index of information quantity. Besides, we experimently show that the obvious correlation between the entropy of the feature map and the corresponding filter, so as to propose an interpretable, data-driven scheme to measure the information capacity of the filter. Further, we introduce information independence, another metric that represents the correlation among differrent filters. Consequently, the least impotant filters, which have less information capacity and less information independence, will be pruned. We evaluate our method on two benchmarks using multiple representative CNN architectures, including VGG-16 and ResNet. On CIFAR-10, we reduce 71.9% of floating-point operations (FLOPs) and 69.4% of parameters for ResNet-110 with 0.28% accuracy increase. On ILSVRC-2012, we reduce 76.6% of floating-point operations (FLOPs) and 68.6% of parameters for ResNet-50 with only 2.80% accuracy decrease, which outperforms the state-of-the-arts.
comment: 9 pages, 5 figures
☆ Sketch-based Medical Image Retrieval
The amount of medical images stored in hospitals is increasing faster than ever; however, utilizing the accumulated medical images has been limited. This is because existing content-based medical image retrieval (CBMIR) systems usually require example images to construct query vectors; nevertheless, example images cannot always be prepared. Besides, there can be images with rare characteristics that make it difficult to find similar example images, which we call isolated samples. Here, we introduce a novel sketch-based medical image retrieval (SBMIR) system that enables users to find images of interest without example images. The key idea lies in feature decomposition of medical images, whereby the entire feature of a medical image can be decomposed into and reconstructed from normal and abnormal features. By extending this idea, our SBMIR system provides an easy-to-use two-step graphical user interface: users first select a template image to specify a normal feature and then draw a semantic sketch of the disease on the template image to represent an abnormal feature. Subsequently, it integrates the two kinds of input to construct a query vector and retrieves reference images with the closest reference vectors. Using two datasets, ten healthcare professionals with various clinical backgrounds participated in the user test for evaluation. As a result, our SBMIR system enabled users to overcome previous challenges, including image retrieval based on fine-grained image characteristics, image retrieval without example images, and image retrieval for isolated samples. Our SBMIR system achieves flexible medical image retrieval on demand, thereby expanding the utility of medical image databases.
☆ No One Left Behind: Improving the Worst Categories in Long-Tailed Learning
Unlike the case when using a balanced training dataset, the per-class recall (i.e., accuracy) of neural networks trained with an imbalanced dataset are known to vary a lot from category to category. The convention in long-tailed recognition is to manually split all categories into three subsets and report the average accuracy within each subset. We argue that under such an evaluation setting, some categories are inevitably sacrificed. On one hand, focusing on the average accuracy on a balanced test set incurs little penalty even if some worst performing categories have zero accuracy. On the other hand, classes in the "Few" subset do not necessarily perform worse than those in the "Many" or "Medium" subsets. We therefore advocate to focus more on improving the lowest recall among all categories and the harmonic mean of all recall values. Specifically, we propose a simple plug-in method that is applicable to a wide range of methods. By simply re-training the classifier of an existing pre-trained model with our proposed loss function and using an optional ensemble trick that combines the predictions of the two classifiers, we achieve a more uniform distribution of recall values across categories, which leads to a higher harmonic mean accuracy while the (arithmetic) average accuracy is still high. The effectiveness of our method is justified on widely used benchmark datasets.
☆ SGDA: Towards 3D Universal Pulmonary Nodule Detection via Slice Grouped Domain Attention
Lung cancer is the leading cause of cancer death worldwide. The best solution for lung cancer is to diagnose the pulmonary nodules in the early stage, which is usually accomplished with the aid of thoracic computed tomography (CT). As deep learning thrives, convolutional neural networks (CNNs) have been introduced into pulmonary nodule detection to help doctors in this labor-intensive task and demonstrated to be very effective. However, the current pulmonary nodule detection methods are usually domain-specific, and cannot satisfy the requirement of working in diverse real-world scenarios. To address this issue, we propose a slice grouped domain attention (SGDA) module to enhance the generalization capability of the pulmonary nodule detection networks. This attention module works in the axial, coronal, and sagittal directions. In each direction, we divide the input feature into groups, and for each group, we utilize a universal adapter bank to capture the feature subspaces of the domains spanned by all pulmonary nodule datasets. Then the bank outputs are combined from the perspective of domain to modulate the input group. Extensive experiments demonstrate that SGDA enables substantially better multi-domain pulmonary nodule detection performance compared with the state-of-the-art multi-domain learning methods.
comment: Accepted by IEEE/ACM Transactions on Computational Biology and Bioinformatics
☆ FSVVD: A Dataset of Full Scene Volumetric Video
Recent years have witnessed a rapid development of immersive multimedia which bridges the gap between the real world and virtual space. Volumetric videos, as an emerging representative 3D video paradigm that empowers extended reality, stand out to provide unprecedented immersive and interactive video watching experience. Despite the tremendous potential, the research towards 3D volumetric video is still in its infancy, relying on sufficient and complete datasets for further exploration. However, existing related volumetric video datasets mostly only include a single object, lacking details about the scene and the interaction between them. In this paper, we focus on the current most widely used data format, point cloud, and for the first time release a full-scene volumetric video dataset that includes multiple people and their daily activities interacting with the external environments. Comprehensive dataset description and analysis are conducted, with potential usage of this dataset. The dataset and additional tools can be accessed via the following website: https://cuhksz-inml.github.io/full_scene_volumetric_video_dataset/.
comment: Accepted by MMSys'23 Open Dataset and Software Track, A preliminary version. The dataset and additional tools can be accessed via https://cuhksz-inml.github.io/full_scene_volumetric_video_dataset/
☆ Guided Image-to-Image Translation by Discriminator-Generator Communication
The goal of Image-to-image (I2I) translation is to transfer an image from a source domain to a target domain, which has recently drawn increasing attention. One major branch of this research is to formulate I2I translation based on Generative Adversarial Network (GAN). As a zero-sum game, GAN can be reformulated as a Partially-observed Markov Decision Process (POMDP) for generators, where generators cannot access full state information of their environments. This formulation illustrates the information insufficiency in the GAN training. To mitigate this problem, we propose to add a communication channel between discriminators and generators. We explore multiple architecture designs to integrate the communication mechanism into the I2I translation framework. To validate the performance of the proposed approach, we have conducted extensive experiments on various benchmark datasets. The experimental results confirm the superiority of our proposed method.
☆ LoGoNet: Towards Accurate 3D Object Detection with Local-to-Global Cross-Modal Fusion CVPR2023
LiDAR-camera fusion methods have shown impressive performance in 3D object detection. Recent advanced multi-modal methods mainly perform global fusion, where image features and point cloud features are fused across the whole scene. Such practice lacks fine-grained region-level information, yielding suboptimal fusion performance. In this paper, we present the novel Local-to-Global fusion network (LoGoNet), which performs LiDAR-camera fusion at both local and global levels. Concretely, the Global Fusion (GoF) of LoGoNet is built upon previous literature, while we exclusively use point centroids to more precisely represent the position of voxel features, thus achieving better cross-modal alignment. As to the Local Fusion (LoF), we first divide each proposal into uniform grids and then project these grid centers to the images. The image features around the projected grid points are sampled to be fused with position-decorated point cloud features, maximally utilizing the rich contextual information around the proposals. The Feature Dynamic Aggregation (FDA) module is further proposed to achieve information interaction between these locally and globally fused features, thus producing more informative multi-modal features. Extensive experiments on both Waymo Open Dataset (WOD) and KITTI datasets show that LoGoNet outperforms all state-of-the-art 3D detection methods. Notably, LoGoNet ranks 1st on Waymo 3D object detection leaderboard and obtains 81.02 mAPH (L2) detection performance. It is noteworthy that, for the first time, the detection performance on three classes surpasses 80 APH (L2) simultaneously. Code will be available at \url{https://github.com/sankin97/LoGoNet}.
comment: Accepted by CVPR2023
☆ Calibration-free BEV Representation for Infrastructure Perception
Effective BEV object detection on infrastructure can greatly improve traffic scenes understanding and vehicle-toinfrastructure (V2I) cooperative perception. However, cameras installed on infrastructure have various postures, and previous BEV detection methods rely on accurate calibration, which is difficult for practical applications due to inevitable natural factors (e.g., wind and snow). In this paper, we propose a Calibration-free BEV Representation (CBR) network, which achieves 3D detection based on BEV representation without calibration parameters and additional depth supervision. Specifically, we utilize two multi-layer perceptrons for decoupling the features from perspective view to front view and birdeye view under boxes-induced foreground supervision. Then, a cross-view feature fusion module matches features from orthogonal views according to similarity and conducts BEV feature enhancement with front view features. Experimental results on DAIR-V2X demonstrate that CBR achieves acceptable performance without any camera parameters and is naturally not affected by calibration noises. We hope CBR can serve as a baseline for future research addressing practical challenges of infrastructure perception.
☆ CLIP-Layout: Style-Consistent Indoor Scene Synthesis with Semantic Furniture Embedding
Indoor scene synthesis involves automatically picking and placing furniture appropriately on a floor plan, so that the scene looks realistic and is functionally plausible. Such scenes can serve as a home for immersive 3D experiences, or be used to train embodied agents. Existing methods for this task rely on labeled categories of furniture, e.g. bed, chair or table, to generate contextually relevant combinations of furniture. Whether heuristic or learned, these methods ignore instance-level attributes of objects such as color and style, and as a result may produce visually less coherent scenes. In this paper, we introduce an auto-regressive scene model which can output instance-level predictions, making use of general purpose image embedding based on CLIP. This allows us to learn visual correspondences such as matching color and style, and produce more plausible and aesthetically pleasing scenes. Evaluated on the 3D-FRONT dataset, our model achieves SOTA results in scene generation and improves auto-completion metrics by over 50%. Moreover, our embedding-based approach enables zero-shot text-guided scene generation and editing, which easily generalizes to furniture not seen at training time.
☆ Diffusion in the Dark: A Diffusion Model for Low-Light Text Recognition
Images are indispensable for the automation of high-level tasks, such as text recognition. Low-light conditions pose a challenge for these high-level perception stacks, which are often optimized on well-lit, artifact-free images. Reconstruction methods for low-light images can produce well-lit counterparts, but typically at the cost of high-frequency details critical for downstream tasks. We propose Diffusion in the Dark (DiD), a diffusion model for low-light image reconstruction that provides qualitatively competitive reconstructions with that of SOTA, while preserving high-frequency details even in extremely noisy, dark conditions. We demonstrate that DiD, without any task-specific optimization, can outperform SOTA low-light methods in low-light text recognition on real images, bolstering the potential of diffusion models for ill-posed inverse problems.
comment: Project website: https://ccnguyen.github.io/diffusion-in-the-dark/
☆ CUDA: Convolution-based Unlearnable Datasets CVPR 2023
Large-scale training of modern deep learning models heavily relies on publicly available data on the web. This potentially unauthorized usage of online data leads to concerns regarding data privacy. Recent works aim to make unlearnable data for deep learning models by adding small, specially designed noises to tackle this issue. However, these methods are vulnerable to adversarial training (AT) and/or are computationally heavy. In this work, we propose a novel, model-free, Convolution-based Unlearnable DAtaset (CUDA) generation technique. CUDA is generated using controlled class-wise convolutions with filters that are randomly generated via a private key. CUDA encourages the network to learn the relation between filters and labels rather than informative features for classifying the clean data. We develop some theoretical analysis demonstrating that CUDA can successfully poison Gaussian mixture data by reducing the clean data performance of the optimal Bayes classifier. We also empirically demonstrate the effectiveness of CUDA with various datasets (CIFAR-10, CIFAR-100, ImageNet-100, and Tiny-ImageNet), and architectures (ResNet-18, VGG-16, Wide ResNet-34-10, DenseNet-121, DeIT, EfficientNetV2-S, and MobileNetV2). Our experiments show that CUDA is robust to various data augmentations and training approaches such as smoothing, AT with different budgets, transfer learning, and fine-tuning. For instance, training a ResNet-18 on ImageNet-100 CUDA achieves only 8.96$\%$, 40.08$\%$, and 20.58$\%$ clean test accuracies with empirical risk minimization (ERM), $L_{\infty}$ AT, and $L_{2}$ AT, respectively. Here, ERM on the clean training data achieves a clean test accuracy of 80.66$\%$. CUDA exhibits unlearnability effect with ERM even when only a fraction of the training dataset is perturbed. Furthermore, we also show that CUDA is robust to adaptive defenses designed specifically to break it.
comment: CVPR 2023
☆ A Computer Vision Enabled damage detection model with improved YOLOv5 based on Transformer Prediction Head
Objective:Computer vision-based up-to-date accurate damage classification and localization are of decisive importance for infrastructure monitoring, safety, and the serviceability of civil infrastructure. Current state-of-the-art deep learning (DL)-based damage detection models, however, often lack superior feature extraction capability in complex and noisy environments, limiting the development of accurate and reliable object distinction. Method: To this end, we present DenseSPH-YOLOv5, a real-time DL-based high-performance damage detection model where DenseNet blocks have been integrated with the backbone to improve in preserving and reusing critical feature information. Additionally, convolutional block attention modules (CBAM) have been implemented to improve attention performance mechanisms for strong and discriminating deep spatial feature extraction that results in superior detection under various challenging environments. Moreover, additional feature fusion layers and a Swin-Transformer Prediction Head (SPH) have been added leveraging advanced self-attention mechanism for more efficient detection of multiscale object sizes and simultaneously reducing the computational complexity. Results: Evaluating the model performance in large-scale Road Damage Dataset (RDD-2018), at a detection rate of 62.4 FPS, DenseSPH-YOLOv5 obtains a mean average precision (mAP) value of 85.25 %, F1-score of 81.18 %, and precision (P) value of 89.51 % outperforming current state-of-the-art models. Significance: The present research provides an effective and efficient damage localization model addressing the shortcoming of existing DL-based damage detection models by providing highly accurate localized bounding box prediction. Current work constitutes a step towards an accurate and robust automated damage detection system in real-time in-field applications.
☆ PSDNet: Determination of Particle Size Distributions Using Synthetic Soil Images and Convolutional Neural Networks
This project aimed to determine the grain size distribution of granular materials from images using convolutional neural networks. The application of ConvNet and pretrained ConvNet models, including AlexNet, SqueezeNet, GoogLeNet, InceptionV3, DenseNet201, MobileNetV2, ResNet18, ResNet50, ResNet101, Xception, InceptionResNetV2, ShuffleNet, and NASNetMobile was studied. Synthetic images of granular materials created with the discrete element code YADE were used. All the models were trained and verified with grayscale and color band datasets with image sizes ranging from 32 to 160 pixels. The proposed ConvNet model predicts the percentages of mass retained on the finest sieve, coarsest sieve, and all sieves with root-mean-square errors of 1.8 %, 3.3 %, and 2.8 %, respectively, and a coefficient of determination of 0.99. For pretrained networks, root-mean-square errors of 2.4 % and 2.8 % were obtained for the finest sieve with feature extraction and transfer learning models, respectively.
☆ Comparing PSDNet, pretrained networks, and traditional feature extraction for predicting the particle size distribution of granular materials from photographs
This study aims to evaluate PSDNet, a series of convolutional neural networks (ConvNets) trained with photographs to predict the particle size distribution of granular materials. Nine traditional feature extraction methods and 15 pretrained ConvNets were also evaluated and compared. A dataset including 9600 photographs of 15 different granular materials was used. The influence of image size and color band was verified by using six image sizes between 32 and 160 pixels, and both grayscale and color images as PSDNet inputs. In addition to random training, validation, and testing datasets, a material removal method was also used to evaluate the performances of each image analysis method. With this method, each material was successively removed from the training and validation datasets and used as the testing dataset. Results show that a combination of all PSDNet color and grayscale features can lead to a root mean square error (RMSE) on the percentages passing as low as 1.8 % with a random testing dataset and 9.1% with the material removal method. For the random datasets, a combination of all traditional features, and the features extracted from InceptionResNetV2 led to RMSE on the percentages passing of 2.3 and 1.7 %, respectively.
☆ SKGHOI: Spatial-Semantic Knowledge Graph for Human-Object Interaction Detection
Detecting human-object interactions (HOIs) is a challenging problem in computer vision. Existing techniques for HOI detection heavily rely on appearance-based features, which may not capture other essential characteristics for accurate detection. Furthermore, the use of transformer-based models for sentiment representation of human-object pairs can be computationally expensive. To address these challenges, we propose a novel graph-based approach, SKGHOI (Spatial-Semantic Knowledge Graph for Human-Object Interaction Detection), that effectively captures the sentiment representation of HOIs by integrating both spatial and semantic knowledge. In a graph, SKGHOI takes the components of interaction as nodes, and the spatial relationships between them as edges. Our approach employs a spatial encoder and a semantic encoder to extract spatial and semantic information, respectively, and then combines these encodings to create a knowledge graph that captures the sentiment representation of HOIs. Compared to existing techniques, SKGHOI is computationally efficient and allows for the incorporation of prior knowledge, making it practical for use in real-world applications. We demonstrate the effectiveness of our proposed method on the widely-used HICO-DET datasets, where it outperforms existing state-of-the-art graph-based methods by a significant margin. Our results indicate that the SKGHOI approach has the potential to significantly improve the accuracy and efficiency of HOI detection, and we anticipate that it will be of great interest to researchers and practitioners working on this challenging task.
comment: 10 pages, 3 figures, 2 tables
☆ Where We Are and What We're Looking At: Query Based Worldwide Image Geo-localization Using Hierarchies and Scenes CVPR 2023
Determining the exact latitude and longitude that a photo was taken is a useful and widely applicable task, yet it remains exceptionally difficult despite the accelerated progress of other computer vision tasks. Most previous approaches have opted to learn a single representation of query images, which are then classified at different levels of geographic granularity. These approaches fail to exploit the different visual cues that give context to different hierarchies, such as the country, state, and city level. To this end, we introduce an end-to-end transformer-based architecture that exploits the relationship between different geographic levels (which we refer to as hierarchies) and the corresponding visual scene information in an image through hierarchical cross-attention. We achieve this by learning a query for each geographic hierarchy and scene type. Furthermore, we learn a separate representation for different environmental scenes, as different scenes in the same location are often defined by completely different visual features. We achieve state of the art street level accuracy on 4 standard geo-localization datasets : Im2GPS, Im2GPS3k, YFCC4k, and YFCC26k, as well as qualitatively demonstrate how our method learns different representations for different visual hierarchies and scenes, which has not been demonstrated in the previous methods. These previous testing datasets mostly consist of iconic landmarks or images taken from social media, which makes them either a memorization task, or biased towards certain places. To address this issue we introduce a much harder testing dataset, Google-World-Streets-15k, comprised of images taken from Google Streetview covering the whole planet and present state of the art results. Our code will be made available in the camera-ready version.
comment: CVPR 2023
☆ TRACT: Denoising Diffusion Models with Transitive Closure Time-Distillation
Denoising Diffusion models have demonstrated their proficiency for generative sampling. However, generating good samples often requires many iterations. Consequently, techniques such as binary time-distillation (BTD) have been proposed to reduce the number of network calls for a fixed architecture. In this paper, we introduce TRAnsitive Closure Time-distillation (TRACT), a new method that extends BTD. For single step diffusion,TRACT improves FID by up to 2.4x on the same architecture, and achieves new single-step Denoising Diffusion Implicit Models (DDIM) state-of-the-art FID (7.4 for ImageNet64, 3.8 for CIFAR10). Finally we tease apart the method through extended ablations. The PyTorch implementation will be released soon.
☆ A Light-Weight Contrastive Approach for Aligning Human Pose Sequences
We present a simple unsupervised method for learning an encoder mapping short 3D pose sequences into embedding vectors suitable for sequence-to-sequence alignment by dynamic time warping. Training samples consist of temporal windows of frames containing 3D body points such as mocap markers or skeleton joints. A light-weight, 3-layer encoder is trained using a contrastive loss function that encourages embedding vectors of augmented sample pairs to have cosine similarity 1, and similarity 0 with all other samples in a minibatch. When multiple scripted training sequences are available, temporal alignments inferred from an initial round of training are harvested to extract additional, cross-performance match pairs for a second phase of training to refine the encoder. In addition to being simple, the proposed method is fast to train, making it easy to adapt to new data using different marker sets or skeletal joint layouts. Experimental results illustrate ease of use, transferability, and utility of the learned embeddings for comparing and analyzing human behavior sequences.
☆ Gradient-Guided Knowledge Distillation for Object Detectors
Deep learning models have demonstrated remarkable success in object detection, yet their complexity and computational intensity pose a barrier to deploying them in real-world applications (e.g., self-driving perception). Knowledge Distillation (KD) is an effective way to derive efficient models. However, only a small number of KD methods tackle object detection. Also, most of them focus on mimicking the plain features of the teacher model but rarely consider how the features contribute to the final detection. In this paper, we propose a novel approach for knowledge distillation in object detection, named Gradient-guided Knowledge Distillation (GKD). Our GKD uses gradient information to identify and assign more weights to features that significantly impact the detection loss, allowing the student to learn the most relevant features from the teacher. Furthermore, we present bounding-box-aware multi-grained feature imitation (BMFI) to further improve the KD performance. Experiments on the KITTI and COCO-Traffic datasets demonstrate our method's efficacy in knowledge distillation for object detection. On one-stage and two-stage detectors, our GKD-BMFI leads to an average of 5.1% and 3.8% mAP improvement, respectively, beating various state-of-the-art KD methods.
comment: 9 pages include references, 4 figures, and 4 tables
☆ Patch of Invisibility: Naturalistic Black-Box Adversarial Attacks on Object Detectors
Adversarial attacks on deep-learning models have been receiving increased attention in recent years. Work in this area has mostly focused on gradient-based techniques, so-called white-box attacks, wherein the attacker has access to the targeted model's internal parameters; such an assumption is usually unrealistic in the real world. Some attacks additionally use the entire pixel space to fool a given model, which is neither practical nor physical (i.e., real-world). On the contrary, we propose herein a gradient-free method that uses the learned image manifold of a pretrained generative adversarial network (GAN) to generate naturalistic physical adversarial patches for object detectors. We show that our proposed method works both digitally and physically.
☆ EscherNet 101
A deep learning model, EscherNet 101, is constructed to categorize images of 2D periodic patterns into their respective 17 wallpaper groups. Beyond evaluating EscherNet 101 performance by classification rates, at a micro-level we investigate the filters learned at different layers in the network, capable of capturing second-order invariants beyond edge and curvature.
comment: 16 page, 12 figures
☆ Deep hybrid model with satellite imagery: how to combine demand modeling and computer vision for behavior analysis?
Classical demand modeling analyzes travel behavior using only low-dimensional numeric data (i.e. sociodemographics and travel attributes) but not high-dimensional urban imagery. However, travel behavior depends on the factors represented by both numeric data and urban imagery, thus necessitating a synergetic framework to combine them. This study creates a theoretical framework of deep hybrid models with a crossing structure consisting of a mixing operator and a behavioral predictor, thus integrating the numeric and imagery data into a latent space. Empirically, this framework is applied to analyze travel mode choice using the MyDailyTravel Survey from Chicago as the numeric inputs and the satellite images as the imagery inputs. We found that deep hybrid models outperform both the traditional demand models and the recent deep learning in predicting the aggregate and disaggregate travel behavior with our supervision-as-mixing design. The latent space in deep hybrid models can be interpreted, because it reveals meaningful spatial and social patterns. The deep hybrid models can also generate new urban images that do not exist in reality and interpret them with economic theory, such as computing substitution patterns and social welfare changes. Overall, the deep hybrid models demonstrate the complementarity between the low-dimensional numeric and high-dimensional imagery data and between the traditional demand modeling and recent deep learning. It generalizes the latent classes and variables in classical hybrid demand models to a latent space, and leverages the computational power of deep learning for imagery while retaining the economic interpretability on the microeconomics foundation.
☆ Toward a Geometric Theory of Manifold Untangling
It has been hypothesized that the ventral stream processing for object recognition is based on a mechanism called cortically local subspace untangling. A mathematical abstraction of object recognition by the visual cortex is how to untangle the manifolds associated with different object category. Such a manifold untangling problem is closely related to the celebrated kernel trick in metric space. In this paper, we conjecture that there is a more general solution to manifold untangling in the topological space without artificially defining any distance metric. Geometrically, we can either $embed$ a manifold in a higher dimensional space to promote selectivity or $flatten$ a manifold to promote tolerance. General strategies of both global manifold embedding and local manifold flattening are presented and connected with existing work on the untangling of image, audio, and language data. We also discuss the implications of untangling the manifold into motor control and internal representations.
☆ Clustering large 3D volumes: A sampling-based approach
In many applications of X-ray computed tomography, an unsupervised segmentation of the reconstructed 3D volumes forms an important step in the image processing chain for further investigation of the digitized object. Therefore, the goal is to train a clustering algorithm on the volume, which produces a voxelwise classification by assigning a cluster index to each voxel. However, clustering methods, e.g., K-Means, typically have an asymptotic polynomial runtime with respect to the dataset size, and thus, these techniques are rarely applicable to large volumes. In this work, we introduce a novel clustering technique based on random sampling, which allows for the voxelwise classification of arbitrarily large volumes. The presented method conducts efficient linear passes over the data to extract a representative random sample of a fixed size on which the classifier can be trained. Then, a final linear pass performs the segmentation and assigns a cluster index to each individual voxel. Quantitative and qualitative evaluations show that excellent results can be achieved even with a very small sample size. Consequently, the unsupervised segmentation by means of clustering becomes feasible for arbitrarily large volumes.
comment: 12 pages, 8 figures
☆ End-to-end Face-swapping via Adaptive Latent Representation Learning
Taking full advantage of the excellent performance of StyleGAN, style transfer-based face swapping methods have been extensively investigated recently. However, these studies require separate face segmentation and blending modules for successful face swapping, and the fixed selection of the manipulated latent code in these works is reckless, thus degrading face swapping quality, generalizability, and practicability. This paper proposes a novel and end-to-end integrated framework for high resolution and attribute preservation face swapping via Adaptive Latent Representation Learning. Specifically, we first design a multi-task dual-space face encoder by sharing the underlying feature extraction network to simultaneously complete the facial region perception and face encoding. This encoder enables us to control the face pose and attribute individually, thus enhancing the face swapping quality. Next, we propose an adaptive latent codes swapping module to adaptively learn the mapping between the facial attributes and the latent codes and select effective latent codes for improved retention of facial attributes. Finally, the initial face swapping image generated by StyleGAN2 is blended with the facial region mask generated by our encoder to address the background blur problem. Our framework integrating facial perceiving and blending into the end-to-end training and testing process can achieve high realistic face-swapping on wild faces without segmentation masks. Experimental results demonstrate the superior performance of our approach over state-of-the-art methods.
☆ Robustness-preserving Lifelong Learning via Dataset Condensation ICASSP2023
Lifelong learning (LL) aims to improve a predictive model as the data source evolves continuously. Most work in this learning paradigm has focused on resolving the problem of 'catastrophic forgetting,' which refers to a notorious dilemma between improving model accuracy over new data and retaining accuracy over previous data. Yet, it is also known that machine learning (ML) models can be vulnerable in the sense that tiny, adversarial input perturbations can deceive the models into producing erroneous predictions. This motivates the research objective of this paper - specification of a new LL framework that can salvage model robustness (against adversarial attacks) from catastrophic forgetting. Specifically, we propose a new memory-replay LL strategy that leverages modern bi-level optimization techniques to determine the 'coreset' of the current data (i.e., a small amount of data to be memorized) for ease of preserving adversarial robustness over time. We term the resulting LL framework 'Data-Efficient Robustness-Preserving LL' (DERPLL). The effectiveness of DERPLL is evaluated for class-incremental image classification using ResNet-18 over the CIFAR-10 dataset. Experimental results show that DERPLL outperforms the conventional coreset-guided LL baseline and achieves a substantial improvement in both standard accuracy and robust accuracy.
comment: Accepted by ICASSP2023 Main Track: Machine Learning for Signal Processing
♻ ☆ TTCDist: Fast Distance Estimation From an Active Monocular Camera Using Time-to-Contact ICRA 2023
Distance estimation from vision is fundamental for a myriad of robotic applications such as navigation, manipulation, and planning. Inspired by the mammal's visual system, which gazes at specific objects, we develop two novel constraints relating time-to-contact, acceleration, and distance that we call the $\tau$-constraint and $\Phi$-constraint. They allow an active (moving) camera to estimate depth efficiently and accurately while using only a small portion of the image. The constraints are applicable to range sensing, sensor fusion, and visual servoing. We successfully validate the proposed constraints with two experiments. The first applies both constraints in a trajectory estimation task with a monocular camera and an Inertial Measurement Unit (IMU). Our methods achieve 30-70% less average trajectory error while running 25$\times$ and 6.2$\times$ faster than the popular Visual-Inertial Odometry methods VINS-Mono and ROVIO respectively. The second experiment demonstrates that when the constraints are used for feedback with efference copies the resulting closed loop system's eigenvalues are invariant to scaling of the applied control signal. We believe these results indicate the $\tau$ and $\Phi$ constraint's potential as the basis of robust and efficient algorithms for a multitude of robotic applications.
comment: 19 pages, 24 figures, 1 table. To be published in ICRA 2023
♻ ☆ On Calibrating Semantic Segmentation Models: Analyses and An Algorithm CVPR2023
We study the problem of semantic segmentation calibration. For image classification, lots of existing solutions are proposed to alleviate model miscalibration of confidence. However, to date, confidence calibration research on semantic segmentation is still limited. We provide a systematic study on the calibration of semantic segmentation models and propose a simple yet effective approach. First, we find that model capacity, crop size, multi-scale testing, and prediction correctness have impact on calibration. Among them, prediction correctness, especially misprediction, is more important to miscalibration due to over-confidence. Next, we propose a simple, unifying, and effective approach, namely selective scaling, by separating correct/incorrect prediction for scaling and more focusing on misprediction logit smoothing. Then, we study popular existing calibration methods and compare them with selective scaling on semantic segmentation calibration. We conduct extensive experiments with a variety of benchmarks on both in-domain and domain-shift calibration, and show that selective scaling consistently outperforms other methods.
comment: Accepted to CVPR2023 (8 pages, 4 figures)
♻ ☆ Learning Reward Functions for Robotic Manipulation by Observing Humans
Observing a human demonstrator manipulate objects provides a rich, scalable and inexpensive source of data for learning robotic policies. However, transferring skills from human videos to a robotic manipulator poses several challenges, not least a difference in action and observation spaces. In this work, we use unlabeled videos of humans solving a wide range of manipulation tasks to learn a task-agnostic reward function for robotic manipulation policies. Thanks to the diversity of this training data, the learned reward function sufficiently generalizes to image observations from a previously unseen robot embodiment and environment to provide a meaningful prior for directed exploration in reinforcement learning. We propose two methods for scoring states relative to a goal image: through direct temporal regression, and through distances in an embedding space obtained with time-contrastive learning. By conditioning the function on a goal image, we are able to reuse one model across a variety of tasks. Unlike prior work on leveraging human videos to teach robots, our method, Human Offline Learned Distances (HOLD) requires neither a priori data from the robot environment, nor a set of task-specific human demonstrations, nor a predefined notion of correspondence across morphologies, yet it is able to accelerate training of several manipulation tasks on a simulated robot arm compared to using only a sparse reward obtained from task completion.
♻ ☆ Iterative Patch Selection for High-Resolution Image Recognition ICLR 2023
High-resolution images are prevalent in various applications, such as autonomous driving and computer-aided diagnosis. However, training neural networks on such images is computationally challenging and easily leads to out-of-memory errors even on modern GPUs. We propose a simple method, Iterative Patch Selection (IPS), which decouples the memory usage from the input size and thus enables the processing of arbitrarily large images under tight hardware constraints. IPS achieves this by selecting only the most salient patches, which are then aggregated into a global representation for image recognition. For both patch selection and aggregation, a cross-attention based transformer is introduced, which exhibits a close connection to Multiple Instance Learning. Our method demonstrates strong performance and has wide applicability across different domains, training regimes and image sizes while using minimal accelerator memory. For example, we are able to finetune our model on whole-slide images consisting of up to 250k patches (>16 gigapixels) with only 5 GB of GPU VRAM at a batch size of 16.
comment: Published as a conference paper at ICLR 2023
♻ ☆ Robust affine feature matching via quadratic assignment on Grassmannians
GraNNI (Grassmannians for Nearest Neighbours Identification) a new algorithm to solve the problem of affine registration is proposed. The algorithm is based on the Grassmannian of $k$--dimensional planes in $\mathbb{R}^n$ and minimizing the Frobenius norm between the two elements of the Grassmannian. The Quadratic Assignment Problem (QAP) is used to find the matching. The results of the experiments show that the algorithm is more robust to noise and point discrepancy in point clouds than previous approaches.
comment: 12 pages, 18 figures; GitHub repository at (https://github.com/sashakolpakov/granni)
♻ ☆ Towards Fast Adaptation of Pretrained Contrastive Models for Multi-channel Video-Language Retrieval CVPR 2023
Multi-channel video-language retrieval require models to understand information from different channels (e.g. video$+$question, video$+$speech) to correctly link a video with a textual response or query. Fortunately, contrastive multimodal models have been shown to be highly effective at aligning entities in images/videos and text, e.g., CLIP; text contrastive models have been extensively studied recently for their strong ability of producing discriminative sentence embeddings, e.g., SimCSE. Their abilities are exactly needed by multi-channel video-language retrieval. However, there is not a clear way to quickly adapt these two lines to multi-channel video-language retrieval with limited data and resources. In this paper, we identify a principled model design space with two axes: how to represent videos and how to fuse video and text information. Based on categorization of recent methods, we investigate the options of representing videos using continuous feature vectors or discrete text tokens; for the fusion method, we explore the use of a multimodal transformer or a pretrained contrastive text model. We extensively evaluate the four combinations on five video-language datasets. We surprisingly find that discrete text tokens coupled with a pretrained contrastive text model yields the best performance, which can even outperform state-of-the-art on the iVQA and How2QA datasets without the additional training on millions of video-language data. Further analysis shows that this is because representing videos as text tokens captures the key visual information with text tokens that are naturally aligned with text models and the text models are strong multimodal retriever after the contrastive pretraining process.
comment: To appear in CVPR 2023
♻ ☆ On the Importance of Sign Labeling: The Hamburg Sign Language Notation System Case Study
Labeling is the cornerstone of supervised machine learning, which has been exploited in a plethora of various applications, with sign language recognition being one of them. However, such algorithms must be fed with a huge amount of consistently labeled data during the training process to elaborate a well-generalizing model. In addition, there is a great need for an automated solution that works with any nationally diversified sign language. Although there are language-agnostic transcription systems, such as the Hamburg Sign Language Notation System (HamNoSys) that describe the signer's initial position and body movement instead of the glosses' meanings, there are still issues with providing accurate and reliable labels for every real-world use case. In this context, the industry relies heavily on manual attribution and labeling of the available video data. In this work, we tackle this issue and thoroughly analyze the HamNoSys labels provided by various maintainers of open sign language corpora in five sign languages, in order to examine the challenges encountered in labeling video data. We also investigate the consistency and objectivity of HamNoSys-based labels for the purpose of training machine learning models. Our findings provide valuable insights into the limitations of the current labeling methods and pave the way for future research on developing more accurate and efficient solutions for sign language recognition.
comment: 20 pages, 13 figures
♻ ☆ Surround-View Vision-based 3D Detection for Autonomous Driving: A Survey
Vision-based 3D Detection task is fundamental task for the perception of an autonomous driving system, which has peaked interest amongst many researchers and autonomous driving engineers. However achieving a rather good 3D BEV (Bird's Eye View) performance is not an easy task using 2D sensor input-data with cameras. In this paper we provide a literature survey for the existing Vision Based 3D detection methods, focused on autonomous driving. We have made detailed analysis of over $60$ papers leveraging Vision BEV detections approaches and highlighted different sub-groups for detailed understanding of common trends. Moreover, we have highlighted how the literature and industry trend have moved towards surround-view image based methods and note down thoughts on what special cases this method addresses. In conclusion, we provoke thoughts of 3D Vision techniques for future research based on shortcomings of the current techniques including the direction of collaborative perception.
♻ ☆ 3M3D: Multi-view, Multi-path, Multi-representation for 3D Object Detection
3D visual perception tasks based on multi-camera images are essential for autonomous driving systems. Latest work in this field performs 3D object detection by leveraging multi-view images as an input and iteratively enhancing object queries (object proposals) by cross-attending multi-view features. However, individual backbone features are not updated with multi-view features and it stays as a mere collection of the output of the single-image backbone network. Therefore we propose 3M3D: A Multi-view, Multi-path, Multi-representation for 3D Object Detection where we update both multi-view features and query features to enhance the representation of the scene in both fine panoramic view and coarse global view. Firstly, we update multi-view features by multi-view axis self-attention. It will incorporate panoramic information in the multi-view features and enhance understanding of the global scene. Secondly, we update multi-view features by self-attention of the ROI (Region of Interest) windows which encodes local finer details in the features. It will help exchange the information not only along the multi-view axis but also along the other spatial dimension. Lastly, we leverage the fact of multi-representation of queries in different domains to further boost the performance. Here we use sparse floating queries along with dense BEV (Bird's Eye View) queries, which are later post-processed to filter duplicate detections. Moreover, we show performance improvements on nuScenes benchmark dataset on top of our baselines.
♻ ☆ Unifying Remote Sensing Image Retrieval and Classification with Robust Fine-tuning
Advances in high resolution remote sensing image analysis are currently hampered by the difficulty of gathering enough annotated data for training deep learning methods, giving rise to a variety of small datasets and associated dataset-specific methods. Moreover, typical tasks such as classification and retrieval lack a systematic evaluation on standard benchmarks and training datasets, which make it hard to identify durable and generalizable scientific contributions. We aim at unifying remote sensing image retrieval and classification with a new large-scale training and testing dataset, SF300, including both vertical and oblique aerial images and made available to the research community, and an associated fine-tuning method. We additionally propose a new adversarial fine-tuning method for global descriptors. We show that our framework systematically achieves a boost of retrieval and classification performance on nine different datasets compared to an ImageNet pretrained baseline, with currently no other method to compare to.
comment: Performance margin with the proposed method is not statistically significant. Please refer to http://alegoria.ign.fr/en/SF300_dataset if you are interested in the dataset
♻ ☆ Aggregation of Disentanglement: Reconsidering Domain Variations in Domain Generalization
Domain Generalization (DG) is a fundamental challenge for machine learning models, which aims to improve model generalization on various domains. Previous methods focus on generating domain invariant features from various source domains. However, we argue that the domain variantions also contain useful information, ie, classification-aware information, for downstream tasks, which has been largely ignored. Different from learning domain invariant features from source domains, we decouple the input images into Domain Expert Features and noise. The proposed domain expert features lie in a learned latent space where the images in each domain can be classified independently, enabling the implicit use of classification-aware domain variations. Based on the analysis, we proposed a novel paradigm called Domain Disentanglement Network (DDN) to disentangle the domain expert features from the source domain images and aggregate the source domain expert features for representing the target test domain. We also propound a new contrastive learning method to guide the domain expert features to form a more balanced and separable feature space. Experiments on the widely-used benchmarks of PACS, VLCS, OfficeHome, DomainNet, and TerraIncognita demonstrate the competitive performance of our method compared to the recently proposed alternatives.
♻ ☆ Efficient Large-scale Scene Representation with a Hybrid of High-resolution Grid and Plane Features
Existing neural radiance fields (NeRF) methods for large-scale scene modeling require days of training using multiple GPUs, hindering their applications in scenarios with limited computing resources. Despite fast optimization NeRF variants have been proposed based on the explicit dense or hash grid features, their effectivenesses are mainly demonstrated in object-scale scene representation. In this paper, we point out that the low feature resolution in explicit representation is the bottleneck for large-scale unbounded scene representation. To address this problem, we introduce a new and efficient hybrid feature representation for NeRF that fuses the 3D hash-grids and high-resolution 2D dense plane features. Compared with the dense-grid representation, the resolution of a dense 2D plane can be scaled up more efficiently. Based on this hybrid representation, we propose a fast optimization NeRF variant, called GP-NeRF, that achieves better rendering results while maintaining a compact model size. Extensive experiments on multiple large-scale unbounded scene datasets show that our model can converge in 1.5 hours using a single GPU while achieving results comparable to or even better than the existing method that requires about one day's training with 8 GPUs.
♻ ☆ NBV-SC: Next Best View Planning based on Shape Completion for Fruit Mapping and Reconstruction
Active perception for fruit mapping and harvesting is a difficult task since occlusions occur frequently and the location as well as size of fruits change over time. State-of-the-art viewpoint planning approaches utilize computationally expensive ray casting operations to find good viewpoints aiming at maximizing information gain and covering the fruits in the scene. In this paper, we present a novel viewpoint planning approach that explicitly uses information about the predicted fruit shapes to compute targeted viewpoints that observe as yet unobserved parts of the fruits. Furthermore, we formulate the concept of viewpoint dissimilarity to reduce the sampling space for more efficient selection of useful, dissimilar viewpoints. Our simulation experiments with a UR5e arm equipped with an RGB-D sensor provide a quantitative demonstration of the efficacy of our iterative next best view planning method based on shape completion. In comparative experiments with a state-of-the-art viewpoint planner, we demonstrate improvement not only in the estimation of the fruit sizes, but also in their reconstruction, while significantly reducing the planning time. Finally, we show the viability of our approach for mapping sweet peppers plants with a real robotic system in a commercial glasshouse.
comment: Agricultural Automation, Viewpoint Planning, Active Perception, Shape Completion
♻ ☆ Early or Late Fusion Matters: Efficient RGB-D Fusion in Vision Transformers for 3D Object Recognition IROS 23
The Vision Transformer (ViT) architecture has established its place in computer vision literature, however, training ViTs for RGB-D object recognition remains an understudied topic, viewed in recent literature only through the lens of multi-task pretraining in multiple vision modalities. Such approaches are often computationally intensive, relying on the scale of multiple pretraining datasets to align RGB with 3D information. In this work, we propose a simple yet strong recipe for transferring pretrained ViTs in RGB-D domains for 3D object recognition, focusing on fusing RGB and depth representations encoded jointly by the ViT. Compared to previous works in multimodal Transformers, the key challenge here is to use the attested flexibility of ViTs to capture cross-modal interactions at the downstream and not the pretraining stage. We explore which depth representation is better in terms of resulting accuracy and compare early and late fusion techniques for aligning the RGB and depth modalities within the ViT architecture. Experimental results in the Washington RGB-D Objects dataset (ROD) demonstrate that in such RGB -> RGB-D scenarios, late fusion techniques work better than most popularly employed early fusion. With our transfer baseline, fusion ViTs score up to 95.4% top-1 accuracy in ROD, achieving new state-of-the-art results in this benchmark. We further show the benefits of using our multimodal fusion baseline over unimodal feature extractors in a synthetic-to-real visual adaptation as well as in an open-ended lifelong learning scenario in the ROD benchmark, where our model outperforms previous works by a margin of >8%. Finally, we integrate our method with a robot framework and demonstrate how it can serve as a perception utility in an interactive robot learning scenario, both in simulation and with a real robot.
comment: Submitted IROS 23. Supplementary video here: https://youtu.be/L2gkDPkHsfo
♻ ☆ Deep Learning for Predictive Analytics in Reversible Steganography
Deep learning is regarded as a promising solution for reversible steganography. There is an accelerating trend of representing a reversible steo-system by monolithic neural networks, which bypass intermediate operations in traditional pipelines of reversible steganography. This end-to-end paradigm, however, suffers from imperfect reversibility. By contrast, the modular paradigm that incorporates neural networks into modules of traditional pipelines can stably guarantee reversibility with mathematical explainability. Prediction-error modulation is a well-established reversible steganography pipeline for digital images. It consists of a predictive analytics module and a reversible coding module. Given that reversibility is governed independently by the coding module, we narrow our focus to the incorporation of neural networks into the analytics module, which serves the purpose of predicting pixel intensities and a pivotal role in determining capacity and imperceptibility. The objective of this study is to evaluate the impacts of different training configurations upon predictive accuracy of neural networks and provide practical insights. In particular, we investigate how different initialisation strategies for input images may affect the learning process and how different training strategies for dual-layer prediction respond to the problem of distributional shift. Furthermore, we compare steganographic performance of various model architectures with different loss functions.
♻ ☆ Positive unlabeled learning with tensor networks
Positive unlabeled learning is a binary classification problem with positive and unlabeled data. It is common in domains where negative labels are costly or impossible to obtain, e.g., medicine and personalized advertising. We apply the locally purified state tensor network to the positive unlabeled learning problem and test our model on the MNIST image and 15 categorical/mixed datasets. On the MNIST dataset, we obtain close to the state-of-the-art results even with very few labeled positive samples. We significantly improve the state-of-the-art on categorical datasets. Further, we show that the agreement fraction between outputs of different models on unlabeled samples is a good indicator of the model's performance. Finally, our method can generate new positive and negative instances, which we demonstrate on simple synthetic datasets.
comment: 12 pages, 5 figures, 4 tables
♻ ☆ Guided Depth Map Super-resolution: A Survey
Guided depth map super-resolution (GDSR), which aims to reconstruct a high-resolution (HR) depth map from a low-resolution (LR) observation with the help of a paired HR color image, is a longstanding and fundamental problem, it has attracted considerable attention from computer vision and image processing communities. A myriad of novel and effective approaches have been proposed recently, especially with powerful deep learning techniques. This survey is an effort to present a comprehensive survey of recent progress in GDSR. We start by summarizing the problem of GDSR and explaining why it is challenging. Next, we introduce some commonly used datasets and image quality assessment methods. In addition, we roughly classify existing GDSR methods into three categories, i.e., filtering-based methods, prior-based methods, and learning-based methods. In each category, we introduce the general description of the published algorithms and design principles, summarize the representative methods, and discuss their highlights and limitations. Moreover, the depth related applications are introduced. Furthermore, we conduct experiments to evaluate the performance of some representative methods based on unified experimental configurations, so as to offer a systematic and fair performance evaluation to readers. Finally, we conclude this survey with possible directions and open problems for further research. All the related materials can be found at \url{https://github.com/zhwzhong/Guided-Depth-Map-Super-resolution-A-Survey}.
comment: Accepted by ACM Computing Surveys
♻ ☆ MOISST: Multi-modal Optimization of Implicit Scene for SpatioTemporal calibration
With the recent advances in autonomous driving and the decreasing cost of LiDARs, the use of multi-modal sensor systems is on the rise. However, in order to make use of the information provided by a variety of complimentary sensors, it is necessary to accurately calibrate them. We take advantage of recent advances in computer graphics and implicit volumetric scene representation to tackle the problem of multi-sensor spatial and temporal calibration. Thanks to a new formulation of the implicit model optimization, we are able to jointly optimize calibration parameters along with scene representation based on radiometric and geometric measurements. Our method enables accurate and robust calibration from data captured in uncontrolled and unstructured urban environments, making our solution more scalable than existing calibration solutions. We demonstrate the accuracy and robustness of our method in urban scenes typically encountered in autonomous driving scenarios.
comment: Project site: https://qherau.github.io/MOISST/
♻ ☆ Denoising Masked AutoEncoders Help Robust Classification ICLR 2023
In this paper, we propose a new self-supervised method, which is called Denoising Masked AutoEncoders (DMAE), for learning certified robust classifiers of images. In DMAE, we corrupt each image by adding Gaussian noises to each pixel value and randomly masking several patches. A Transformer-based encoder-decoder model is then trained to reconstruct the original image from the corrupted one. In this learning paradigm, the encoder will learn to capture relevant semantics for the downstream tasks, which is also robust to Gaussian additive noises. We show that the pre-trained encoder can naturally be used as the base classifier in Gaussian smoothed models, where we can analytically compute the certified radius for any data point. Although the proposed method is simple, it yields significant performance improvement in downstream classification tasks. We show that the DMAE ViT-Base model, which just uses 1/10 parameters of the model developed in recent work arXiv:2206.10550, achieves competitive or better certified accuracy in various settings. The DMAE ViT-Large model significantly surpasses all previous results, establishing a new state-of-the-art on ImageNet dataset. We further demonstrate that the pre-trained model has good transferability to the CIFAR-10 dataset, suggesting its wide adaptability. Models and code are available at https://github.com/quanlin-wu/dmae.
comment: ICLR 2023
♻ ☆ Bayesian Neural Networks for Reversible Steganography
Recent advances in deep learning have led to a paradigm shift in the field of reversible steganography. A fundamental pillar of reversible steganography is predictive modelling which can be realised via deep neural networks. However, non-trivial errors exist in inferences about some out-of-distribution and noisy data. In view of this issue, we propose to consider uncertainty in predictive models based upon a theoretical framework of Bayesian deep learning, thereby creating an adaptive steganographic system. Most modern deep-learning models are regarded as deterministic because they only offer predictions while failing to provide uncertainty measurement. Bayesian neural networks bring a probabilistic perspective to deep learning and can be regarded as self-aware intelligent machinery; that is, a machine that knows its own limitations. To quantify uncertainty, we apply Bayesian statistics to model the predictive distribution and approximate it through Monte Carlo sampling with stochastic forward passes. We further show that predictive uncertainty can be disentangled into aleatoric and epistemic uncertainties and these quantities can be learnt unsupervised. Experimental results demonstrate an improvement delivered by Bayesian uncertainty analysis upon steganographic rate-distortion performance.
♻ ☆ On the predictability in reversible steganography
Artificial neural networks have advanced the frontiers of reversible steganography. The core strength of neural networks is the ability to render accurate predictions for a bewildering variety of data. Residual modulation is recognised as the most advanced reversible steganographic algorithm for digital images. The pivot of this algorithm is predictive analytics in which pixel intensities are predicted given some pixel-wise contextual information. This task can be perceived as a low-level vision problem and hence neural networks for addressing a similar class of problems can be deployed. On top of the prior art, this paper investigates predictability of pixel intensities based on supervised and unsupervised learning frameworks. Predictability analysis enables adaptive data embedding, which in turn leads to a better trade-off between capacity and imperceptibility. While conventional methods estimate predictability by the statistics of local image patterns, learning-based frameworks consider further the degree to which correct predictions can be made by a designated predictor. Not only should the image patterns be taken into account but also the predictor in use. Experimental results show that steganographic performance can be significantly improved by incorporating the learning-based predictability analysers into a reversible steganographic system.
♻ ☆ Blur Interpolation Transformer for Real-World Motion from Blur CVPR2023
This paper studies the challenging problem of recovering motion from blur, also known as joint deblurring and interpolation or blur temporal super-resolution. The challenges are twofold: 1) the current methods still leave considerable room for improvement in terms of visual quality even on the synthetic dataset, and 2) poor generalization to real-world data. To this end, we propose a blur interpolation transformer (BiT) to effectively unravel the underlying temporal correlation encoded in blur. Based on multi-scale residual Swin transformer blocks, we introduce dual-end temporal supervision and temporally symmetric ensembling strategies to generate effective features for time-varying motion rendering. In addition, we design a hybrid camera system to collect the first real-world dataset of one-to-many blur-sharp video pairs. Experimental results show that BiT has a significant gain over the state-of-the-art methods on the public dataset Adobe240. Besides, the proposed real-world dataset effectively helps the model generalize well to real blurry scenarios. Code and data are available at https://github.com/zzh-tech/BiT.
comment: Accepted by CVPR2023
♻ ☆ Semi-Sparsity for Smoothing Filters
In this paper, we propose an interesting semi-sparsity smoothing algorithm based on a novel sparsity-inducing optimization framework. This method is derived from the multiple observations that semi-sparsity prior knowledge is more universally applicable, especially in areas where sparsity is not fully admitted, such as polynomial-smoothing surfaces. We illustrate that this semi-sparsity can be identified into a generalized $L_0$-norm minimization in higher-order gradient domains, thereby giving rise to a new "feature-aware" filtering method with a powerful simultaneous-fitting ability in both sparse features (singularities and sharpening edges) and non-sparse regions (polynomial-smoothing surfaces). Notice that a direct solver is always unavailable due to the non-convexity and combinatorial nature of $L_0$-norm minimization. Instead, we solve the model based on an efficient half-quadratic splitting minimization with fast Fourier transforms (FFTs) for acceleration. We finally demonstrate its versatility and many benefits to a series of signal/image processing and computer vision applications.
comment: Final version but delete the graphic processing part
♻ ☆ Private, fair and accurate: Training large-scale, privacy-preserving AI models in medical imaging
Artificial intelligence (AI) models are increasingly used in the medical domain. However, as medical data is highly sensitive, special precautions to ensure its protection are required. The gold standard for privacy preservation is the introduction of differential privacy (DP) to model training. Prior work indicates that DP has negative implications on model accuracy and fairness, which are unacceptable in medicine and represent a main barrier to the widespread use of privacy-preserving techniques. In this work, we evaluated the effect of privacy-preserving training of AI models for chest radiograph diagnosis regarding accuracy and fairness compared to non-private training. For this, we used a large dataset (N=193,311) of high quality clinical chest radiographs, which were retrospectively collected and manually labeled by experienced radiologists. We then compared non-private deep convolutional neural networks (CNNs) and privacy-preserving (DP) models with respect to privacy-utility trade-offs measured as area under the receiver-operator-characteristic curve (AUROC), and privacy-fairness trade-offs, measured as Pearson's r or Statistical Parity Difference. We found that the non-private CNNs achieved an average AUROC score of 0.90 +- 0.04 over all labels, whereas the DP CNNs with a privacy budget of epsilon=7.89 resulted in an AUROC of 0.87 +- 0.04, i.e., a mere 2.6% performance decrease compared to non-private training. Furthermore, we found the privacy-preserving training not to amplify discrimination against age, sex or co-morbidity. Our study shows that -- under the challenging realistic circumstances of a real-life clinical dataset -- the privacy-preserving training of diagnostic deep learning models is possible with excellent diagnostic accuracy and fairness.
comment: 3 tables, 5 figures, 11 supplementary materials
♻ ☆ Q-Net: Query-Informed Few-Shot Medical Image Segmentation
Deep learning has achieved tremendous success in computer vision, while medical image segmentation (MIS) remains a challenge, due to the scarcity of data annotations. Meta-learning techniques for few-shot segmentation (Meta-FSS) have been widely used to tackle this challenge, while they neglect possible distribution shifts between the query image and the support set. In contrast, an experienced clinician can perceive and address such shifts by borrowing information from the query image, then fine-tune or calibrate her prior cognitive model accordingly. Inspired by this, we propose Q-Net, a Query-informed Meta-FSS approach, which mimics in spirit the learning mechanism of an expert clinician. We build Q-Net based on ADNet, a recently proposed anomaly detection-inspired method. Specifically, we add two query-informed computation modules into ADNet, namely a query-informed threshold adaptation module and a query-informed prototype refinement module. Combining them with a dual-path extension of the feature extraction module, Q-Net achieves state-of-the-art performance on widely used abdominal and cardiac magnetic resonance (MR) image datasets. Our work sheds light on a novel way to improve Meta-FSS techniques by leveraging query information.
comment: Accpeted by Intelligent Systems Conference (IntelliSys) 2023
♻ ☆ DRO: Deep Recurrent Optimizer for Video to Depth
There are increasing interests of studying the video-to-depth (V2D) problem with machine learning techniques. While earlier methods directly learn a mapping from images to depth maps and camera poses, more recent works enforce multi-view geometry constraints through optimization embedded in the learning framework. This paper presents a novel optimization method based on recurrent neural networks to further exploit the potential of neural networks in V2D. Specifically, our neural optimizer alternately updates the depth and camera poses through iterations to minimize a feature-metric cost, and two gated recurrent units iteratively improve the results by tracing historical information. Extensive experimental results demonstrate that our method outperforms previous methods and is more efficient in computation and memory consumption than cost-volume-based methods. In particular, our self-supervised method outperforms previous supervised methods on the KITTI and ScanNet datasets. Our source code is available at https://github.com/aliyun/dro-sfm.
comment: Accepted by IEEE Robotics and Automation Letters
♻ ☆ Exploring Video Quality Assessment on User Generated Contents from Aesthetic and Technical Perspectives
The rapid increase in user-generated-content (UGC) videos calls for the development of effective video quality assessment (VQA) algorithms. However, the objective of the UGC-VQA problem is still ambiguous and can be viewed from two perspectives: the technical perspective, measuring the perception of distortions; and the aesthetic perspective, which relates to preference and recommendation on contents. To understand how these two perspectives affect overall subjective opinions in UGC-VQA, we conduct a large-scale subjective study to collect human quality opinions on overall quality of videos as well as perceptions from aesthetic and technical perspectives. The collected Disentangled Video Quality Database (DIVIDE-3k) confirms that human quality opinions on UGC videos are universally and inevitably affected by both aesthetic and technical perspectives. In light of this, we propose the Disentangled Objective Video Quality Evaluator (DOVER) to learn the quality of UGC videos based on the two perspectives. The DOVER proves state-of-the-art performance in UGC-VQA under very high efficiency. With perspective opinions in DIVIDE-3k, we further propose DOVER++, the first approach to provide reliable clear-cut quality evaluations from a single aesthetic or technical perspective. Code at https://github.com/VQAssessment/DOVER.
♻ ☆ GraspNeRF: Multiview-based 6-DoF Grasp Detection for Transparent and Specular Objects Using Generalizable NeRF ICRA
In this work, we tackle 6-DoF grasp detection for transparent and specular objects, which is an important yet challenging problem in vision-based robotic systems, due to the failure of depth cameras in sensing their geometry. We, for the first time, propose a multiview RGB-based 6-DoF grasp detection network, GraspNeRF, that leverages the generalizable neural radiance field (NeRF) to achieve material-agnostic object grasping in clutter. Compared to the existing NeRF-based 3-DoF grasp detection methods that rely on densely captured input images and time-consuming per-scene optimization, our system can perform zero-shot NeRF construction with sparse RGB inputs and reliably detect 6-DoF grasps, both in real-time. The proposed framework jointly learns generalizable NeRF and grasp detection in an end-to-end manner, optimizing the scene representation construction for the grasping. For training data, we generate a large-scale photorealistic domain-randomized synthetic dataset of grasping in cluttered tabletop scenes that enables direct transfer to the real world. Our extensive experiments in synthetic and real-world environments demonstrate that our method significantly outperforms all the baselines in all the experiments while remaining in real-time. Project page can be found at https://pku-epic.github.io/GraspNeRF
comment: IEEE International Conference on Robotics and Automation (ICRA), 2023
♻ ☆ Taming Self-Supervised Learning for Presentation Attack Detection: In-Image De-Folding and Out-of-Image De-Mixing
Biometric systems are vulnerable to Presentation Attacks (PA) performed using various Presentation Attack Instruments (PAIs). Even though there are numerous Presentation Attack Detection (PAD) techniques based on both deep learning and hand-crafted features, the generalization of PAD for unknown PAI is still a challenging problem. In this work, we empirically prove that the initialization of the PAD model is a crucial factor for the generalization, which is rarely discussed in the community. Based on such observation, we proposed a self-supervised learning-based method, denoted as DF-DM. Specifically, DF-DM is based on a global-local view coupled with De-Folding and De-Mixing to derive the task-specific representation for PAD. During De-Folding, the proposed technique will learn region-specific features to represent samples in a local pattern by explicitly minimizing generative loss. While De-Mixing drives detectors to obtain the instance-specific features with global information for more comprehensive representation by minimizing interpolation-based consistency. Extensive experimental results show that the proposed method can achieve significant improvements in terms of both face and fingerprint PAD in more complicated and hybrid datasets when compared with state-of-the-art methods. When training in CASIA-FASD and Idiap Replay-Attack, the proposed method can achieve an 18.60% Equal Error Rate (EER) in OULU-NPU and MSU-MFSD, exceeding baseline performance by 9.54%. The source code of the proposed technique is available at https://github.com/kongzhecn/dfdm.
comment: Accepted by IEEE Transactions on Neural Networks and Learning Systems (TNNLS)
♻ ☆ Data-Driven Target Localization Using Adaptive Radar Processing and Convolutional Neural Networks
Facilitated by the recent emergence of radio frequency (RF) modeling and simulation tools purposed for adaptive radar processing applications, data-driven approaches to classical problems in radar have rapidly grown in popularity over the past decade. Despite this surge, limited focus has been directed toward the theoretical foundations of these data-driven approaches. In this regard, using adaptive radar processing techniques, we propose a data-driven approach in this work to address the classical problem of radar target localization post adaptive radar detection. To give context to the performance of this data-driven approach, we first analyze the asymptotic breakdown signal-to-clutter-plus-noise ratio (SCNR) threshold of the normalized adaptive matched filter (NAMF) test statistic within the context of radar target localization, and augment this analysis through our proposed deep learning framework for target location estimation. In this procedure, we generate comprehensive datasets by randomly placing targets of variable strengths in predetermined constrained areas using RFView, a site-specific, digital twin, RF modeling and simulation tool. For each radar return from these predefined constrained areas, we generate heatmap tensors in range, azimuth, and elevation of the NAMF test statistic, and of the output power of a generalized sidelobe canceller (GSC). Using our deep learning framework, we estimate target locations from these heatmap tensors to demonstrate the feasibility of and significant improvements provided by our data-driven approach across matched and mismatched settings.
comment: 34 pages, 22 figures. Submitted to IEEE Transactions on Aerospace and Electronic Systems
♻ ☆ Fingerprint Presentation Attack Detection by Channel-wise Feature Denoising
Due to the diversity of attack materials, fingerprint recognition systems (AFRSs) are vulnerable to malicious attacks. It is thus important to propose effective fingerprint presentation attack detection (PAD) methods for the safety and reliability of AFRSs. However, current PAD methods often exhibit poor robustness under new attack types settings. This paper thus proposes a novel channel-wise feature denoising fingerprint PAD (CFD-PAD) method by handling the redundant noise information ignored in previous studies. The proposed method learns important features of fingerprint images by weighing the importance of each channel and identifying discriminative channels and "noise" channels. Then, the propagation of "noise" channels is suppressed in the feature map to reduce interference. Specifically, a PA-Adaptation loss is designed to constrain the feature distribution to make the feature distribution of live fingerprints more aggregate and that of spoof fingerprints more disperse. Experimental results evaluated on the LivDet 2017 dataset showed that the proposed CFD-PAD can achieve a 2.53% average classification error (ACE) and a 93.83% true detection rate when the false detection rate equals 1.0% (TDR@FDR=1%). Also, the proposed method markedly outperforms the best single-model-based methods in terms of ACE (2.53% vs. 4.56%) and TDR@FDR=1%(93.83% vs. 73.32%), which demonstrates its effectiveness. Although we have achieved a comparable result with the state-of-the-art multiple-model-based methods, there still is an increase in TDR@FDR=1% from 91.19% to 93.83%. In addition, the proposed model is simpler, lighter and more efficient and has achieved a 74.76% reduction in computation time compared with the state-of-the-art multiple-model-based method. The source code is available at https://github.com/kongzhecn/cfd-pad.
comment: 15 pages, 8 figures, Accepted by TIFS
♻ ☆ Action-GPT: Leveraging Large-scale Language Models for Improved and Generalized Action Generation
We introduce Action-GPT, a plug-and-play framework for incorporating Large Language Models (LLMs) into text-based action generation models. Action phrases in current motion capture datasets contain minimal and to-the-point information. By carefully crafting prompts for LLMs, we generate richer and fine-grained descriptions of the action. We show that utilizing these detailed descriptions instead of the original action phrases leads to better alignment of text and motion spaces. We introduce a generic approach compatible with stochastic (e.g. VAE-based) and deterministic (e.g. MotionCLIP) text-to-motion models. In addition, the approach enables multiple text descriptions to be utilized. Our experiments show (i) noticeable qualitative and quantitative improvement in the quality of synthesized motions, (ii) benefits of utilizing multiple LLM-generated descriptions, (iii) suitability of the prompt function, and (iv) zero-shot generation capabilities of the proposed approach. Project page: https://actiongpt.github.io
comment: Code, pretrained models and sample videos will be made available at \url{https://actiongpt.github.io}
♻ ☆ Few-shot Image Generation with Diffusion Models
Denoising diffusion probabilistic models (DDPMs) have been proven capable of synthesizing high-quality images with remarkable diversity when trained on large amounts of data. However, to our knowledge, few-shot image generation tasks have yet to be studied with DDPM-based approaches. Modern approaches are mainly built on Generative Adversarial Networks (GANs) and adapt models pre-trained on large source domains to target domains using a few available samples. In this paper, we make the first attempt to study when do DDPMs overfit and suffer severe diversity degradation as training data become scarce. Then we fine-tune DDPMs pre-trained on large source domains to solve the overfitting problem when training data is limited. Although the directly fine-tuned models accelerate convergence and improve generation quality and diversity compared with training from scratch, they still fail to retain some diverse features and can only produce coarse images. Therefore, we design a DDPM pairwise adaptation (DDPM-PA) approach to optimize few-shot DDPM domain adaptation. DDPM-PA efficiently preserves information learned from source domains by keeping the relative pairwise distances between generated samples during adaptation. Besides, DDPM-PA enhances the learning of high-frequency details from source models and limited training data. DDPM-PA further improves generation quality and diversity and achieves results better than current state-of-the-art GAN-based approaches. We demonstrate the effectiveness of our approach on a series of few-shot image generation tasks qualitatively and quantitatively.
comment: Updated Version
♻ ☆ DwinFormer: Dual Window Transformers for End-to-End Monocular Depth Estimation
Depth estimation from a single image is of paramount importance in the realm of computer vision, with a multitude of applications. Conventional methods suffer from the trade-off between consistency and fine-grained details due to the local-receptive field limiting their practicality. This lack of long-range dependency inherently comes from the convolutional neural network part of the architecture. In this paper, a dual window transformer-based network, namely DwinFormer, is proposed, which utilizes both local and global features for end-to-end monocular depth estimation. The DwinFormer consists of dual window self-attention and cross-attention transformers, Dwin-SAT and Dwin-CAT, respectively. The Dwin-SAT seamlessly extracts intricate, locally aware features while concurrently capturing global context. It harnesses the power of local and global window attention to adeptly capture both short-range and long-range dependencies, obviating the need for complex and computationally expensive operations, such as attention masking or window shifting. Moreover, Dwin-SAT introduces inductive biases which provide desirable properties, such as translational equvariance and less dependence on large-scale data. Furthermore, conventional decoding methods often rely on skip connections which may result in semantic discrepancies and a lack of global context when fusing encoder and decoder features. In contrast, the Dwin-CAT employs both local and global window cross-attention to seamlessly fuse encoder and decoder features with both fine-grained local and contextually aware global information, effectively amending semantic gap. Empirical evidence obtained through extensive experimentation on the NYU-Depth-V2 and KITTI datasets demonstrates the superiority of the proposed method, consistently outperforming existing approaches across both indoor and outdoor environments.
♻ ☆ Multi-Modal 3D Object Detection in Autonomous Driving: a Survey
In this survey, we first introduce the background of popular sensors used for self-driving, their data properties, and the corresponding object detection algorithms. Next, we discuss existing datasets that can be used for evaluating multi-modal 3D object detection algorithms. Then we present a review of multi-modal fusion based 3D detection networks, taking a close look at their fusion stage, fusion input and fusion granularity, and how these design choices evolve with time and technology. After the review, we discuss open challenges as well as possible solutions. We hope that this survey can help researchers to get familiar with the field and embark on investigations in the area of multi-modal 3D object detection.
comment: Accepted by International Journal of Computer Vision (IJCV)
♻ ☆ DP-NeRF: Deblurred Neural Radiance Field with Physical Scene Priors CVPR 2023
Neural Radiance Field (NeRF) has exhibited outstanding three-dimensional (3D) reconstruction quality via the novel view synthesis from multi-view images and paired calibrated camera parameters. However, previous NeRF-based systems have been demonstrated under strictly controlled settings, with little attention paid to less ideal scenarios, including with the presence of noise such as exposure, illumination changes, and blur. In particular, though blur frequently occurs in real situations, NeRF that can handle blurred images has received little attention. The few studies that have investigated NeRF for blurred images have not considered geometric and appearance consistency in 3D space, which is one of the most important factors in 3D reconstruction. This leads to inconsistency and the degradation of the perceptual quality of the constructed scene. Hence, this paper proposes a DP-NeRF, a novel clean NeRF framework for blurred images, which is constrained with two physical priors. These priors are derived from the actual blurring process during image acquisition by the camera. DP-NeRF proposes rigid blurring kernel to impose 3D consistency utilizing the physical priors and adaptive weight proposal to refine the color composition error in consideration of the relationship between depth and blur. We present extensive experimental results for synthetic and real scenes with two types of blur: camera motion blur and defocus blur. The results demonstrate that DP-NeRF successfully improves the perceptual quality of the constructed NeRF ensuring 3D geometric and appearance consistency. We further demonstrate the effectiveness of our model with comprehensive ablation analysis.
comment: To be appeared at CVPR 2023, Code: https://github.com/dogyoonlee/DP-NeRF, Project page: https://dogyoonlee.github.io/dpnerf/
♻ ☆ Deconstructed Generation-Based Zero-Shot Model AAAI 2023
Recent research on Generalized Zero-Shot Learning (GZSL) has focused primarily on generation-based methods. However, current literature has overlooked the fundamental principles of these methods and has made limited progress in a complex manner. In this paper, we aim to deconstruct the generator-classifier framework and provide guidance for its improvement and extension. We begin by breaking down the generator-learned unseen class distribution into class-level and instance-level distributions. Through our analysis of the role of these two types of distributions in solving the GZSL problem, we generalize the focus of the generation-based approach, emphasizing the importance of (i) attribute generalization in generator learning and (ii) independent classifier learning with partially biased data. We present a simple method based on this analysis that outperforms SotAs on four public GZSL datasets, demonstrating the validity of our deconstruction. Furthermore, our proposed method remains effective even without a generative model, representing a step towards simplifying the generator-classifier structure. Our code is available at \url{https://github.com/cdb342/DGZ}.
comment: AAAI 2023
♻ ☆ DropKey CVPR2023
In this paper, we focus on analyzing and improving the dropout technique for self-attention layers of Vision Transformer, which is important while surprisingly ignored by prior works. In particular, we conduct researches on three core questions: First, what to drop in self-attention layers? Different from dropping attention weights in literature, we propose to move dropout operations forward ahead of attention matrix calculation and set the Key as the dropout unit, yielding a novel dropout-before-softmax scheme. We theoretically verify that this scheme helps keep both regularization and probability features of attention weights, alleviating the overfittings problem to specific patterns and enhancing the model to globally capture vital information; Second, how to schedule the drop ratio in consecutive layers? In contrast to exploit a constant drop ratio for all layers, we present a new decreasing schedule that gradually decreases the drop ratio along the stack of self-attention layers. We experimentally validate the proposed schedule can avoid overfittings in low-level features and missing in high-level semantics, thus improving the robustness and stableness of model training; Third, whether need to perform structured dropout operation as CNN? We attempt patch-based block-version of dropout operation and find that this useful trick for CNN is not essential for ViT. Given exploration on the above three questions, we present the novel DropKey method that regards Key as the drop unit and exploits decreasing schedule for drop ratio, improving ViTs in a general way. Comprehensive experiments demonstrate the effectiveness of DropKey for various ViT architectures, e.g. T2T and VOLO, as well as for various vision tasks, e.g., image classification, object detection, human-object interaction detection and human body shape recovery. Codes will be released upon acceptance.
comment: Accepted by CVPR2023
♻ ☆ Knowledge Restore and Transfer for Multi-label Class-Incremental Learning
Current class-incremental learning research mainly focuses on single-label classification tasks while multi-label class-incremental learning (MLCIL) with more practical application scenarios is rarely studied. Although there have been many anti-forgetting methods to solve the problem of catastrophic forgetting in class-incremental learning, these methods have difficulty in solving the MLCIL problem due to label absence and information dilution. In this paper, we propose a knowledge restore and transfer (KRT) framework for MLCIL, which includes a dynamic pseudo-label (DPL) module to restore the old class knowledge and an incremental cross-attention(ICA) module to save session-specific knowledge and transfer old class knowledge to the new model sufficiently. Besides, we propose a token loss to jointly optimize the incremental cross-attention module. Experimental results on MS-COCO and PASCAL VOC datasets demonstrate the effectiveness of our method for improving recognition performance and mitigating forgetting on multi-label class-incremental learning tasks.
♻ ☆ StyO: Stylize Your Face in Only One-Shot
This paper focuses on face stylization with a single artistic target. Existing works for this task often fail to retain the source content while achieving geometry variation. Here, we present a novel StyO model, ie. Stylize the face in only One-shot, to solve the above problem. In particular, StyO exploits a disentanglement and recombination strategy. It first disentangles the content and style of source and target images into identifiers, which are then recombined in a cross manner to derive the stylized face image. In this way, StyO decomposes complex images into independent and specific attributes, and simplifies one-shot face stylization as the combination of different attributes from input images, thus producing results better matching face geometry of target image and content of source one. StyO is implemented with latent diffusion models (LDM) and composed of two key modules: 1) Identifier Disentanglement Learner (IDL) for disentanglement phase. It represents identifiers as contrastive text prompts, ie. positive and negative descriptions. And it introduces a novel triple reconstruction loss to fine-tune the pre-trained LDM for encoding style and content into corresponding identifiers; 2) Fine-grained Content Controller (FCC) for the recombination phase. It recombines disentangled identifiers from IDL to form an augmented text prompt for generating stylized faces. In addition, FCC also constrains the cross-attention maps of latent and text features to preserve source face details in results. The extensive evaluation shows that StyO produces high-quality images on numerous paintings of various styles and outperforms the current state-of-the-art. Code will be released upon acceptance.
♻ ☆ ESceme: Vision-and-Language Navigation with Episodic Scene Memory
Vision-and-language navigation (VLN) simulates a visual agent that follows natural-language navigation instructions in real-world scenes. Existing approaches have made enormous progress in navigation in new environments, such as beam search, pre-exploration, and dynamic or hierarchical history encoding. To balance generalization and efficiency, we resort to memorizing visited scenarios apart from the ongoing route while navigating. In this work, we introduce a mechanism of Episodic Scene memory (ESceme) for VLN that wakes an agent's memories of past visits when it enters the current scene. The episodic scene memory allows the agent to envision a bigger picture of the next prediction. This way, the agent learns to utilize dynamically updated information instead of merely adapting to static observations. We provide a simple yet effective implementation of ESceme by enhancing the accessible views at each location and progressively completing the memory while navigating. We verify the superiority of ESceme on short-horizon (R2R), long-horizon (R4R), and vision-and-dialog (CVDN) VLN tasks. Our ESceme also wins first place on the CVDN leaderboard. Code is available: \url{https://github.com/qizhust/esceme}.}
comment: Tech. report; typos corrected
♻ ☆ Vision Transformers: From Semantic Segmentation to Dense Prediction CVPR 2021
The emergence of vision transformers (ViTs) in image classification has shifted the methodologies for visual representation learning. In particular, ViTs learn visual representation at full receptive field per layer across all the image patches, in comparison to the increasing receptive fields of CNNs across layers and other alternatives (e.g., large kernels and atrous convolution). In this work, for the first time we explore the global context learning potentials of ViTs for dense visual prediction (e.g., semantic segmentation). Our motivation is that through learning global context at full receptive field layer by layer, ViTs may capture stronger long-range dependency information, critical for dense prediction tasks. We first demonstrate that encoding an image as a sequence of patches, a vanilla ViT without local convolution and resolution reduction can yield stronger visual representation for semantic segmentation. For example, our model, termed as SEgmentation TRansformer (SETR), excels on ADE20K (50.28% mIoU, the first position in the test leaderboard on the day of submission) and Pascal Context (55.83% mIoU), and performs competitively on Cityscapes. For tackling general dense visual prediction tasks in a cost-effective manner, we further formulate a family of Hierarchical Local-Global (HLG) Transformers, characterized by local attention within windows and global-attention across windows in a pyramidal architecture. Extensive experiments show that our methods achieve appealing performance on a variety of dense prediction tasks (e.g., object detection and instance segmentation and semantic segmentation) as well as image classification. Our code and models are available at https://github.com/fudan-zvg/SETR.
comment: Extended version of CVPR 2021 paper arXiv:2012.15840
♻ ☆ Physics-Informed Machine Learning: A Survey on Problems, Methods and Applications
Recent advances of data-driven machine learning have revolutionized fields like computer vision, reinforcement learning, and many scientific and engineering domains. In many real-world and scientific problems, systems that generate data are governed by physical laws. Recent work shows that it provides potential benefits for machine learning models by incorporating the physical prior and collected data, which makes the intersection of machine learning and physics become a prevailing paradigm. By integrating the data and mathematical physics models seamlessly, it can guide the machine learning model towards solutions that are physically plausible, improving accuracy and efficiency even in uncertain and high-dimensional contexts. In this survey, we present this learning paradigm called Physics-Informed Machine Learning (PIML) which is to build a model that leverages empirical data and available physical prior knowledge to improve performance on a set of tasks that involve a physical mechanism. We systematically review the recent development of physics-informed machine learning from three perspectives of machine learning tasks, representation of physical prior, and methods for incorporating physical prior. We also propose several important open research problems based on the current trends in the field. We argue that encoding different forms of physical prior into model architectures, optimizers, inference algorithms, and significant domain-specific applications like inverse engineering design and robotic control is far from being fully explored in the field of physics-informed machine learning. We believe that the interdisciplinary research of physics-informed machine learning will significantly propel research progress, foster the creation of more effective machine learning models, and also offer invaluable assistance in addressing long-standing problems in related disciplines.
♻ ☆ FractalAD: A simple industrial anomaly detection method using fractal anomaly generation and backbone knowledge distillation
Although industrial anomaly detection (AD) technology has made significant progress in recent years, generating realistic anomalies and learning priors of normal remain challenging tasks. In this study, we propose an end-to-end industrial anomaly detection method called FractalAD. Training samples are obtained by synthesizing fractal images and patches from normal samples. This fractal anomaly generation method is designed to sample the full morphology of anomalies. Moreover, we designed a backbone knowledge distillation structure to extract prior knowledge contained in normal samples. The differences between a teacher and a student model are converted into anomaly attention using a cosine similarity attention module. The proposed method enables an end-to-end semantic segmentation network to be used for anomaly detection without adding any trainable parameters to the backbone and segmentation head, and has obvious advantages over other methods in training and inference speed.. The results of ablation studies confirmed the effectiveness of fractal anomaly generation and backbone knowledge distillation. The results of performance experiments showed that FractalAD achieved competitive results on the MVTec AD dataset and MVTec 3D-AD dataset compared with other state-of-the-art anomaly detection methods.
comment: 12 pages, 5 figures
♻ ☆ VIP: Towards Universal Visual Reward and Representation via Value-Implicit Pre-Training ICLR 2023
Reward and representation learning are two long-standing challenges for learning an expanding set of robot manipulation skills from sensory observations. Given the inherent cost and scarcity of in-domain, task-specific robot data, learning from large, diverse, offline human videos has emerged as a promising path towards acquiring a generally useful visual representation for control; however, how these human videos can be used for general-purpose reward learning remains an open question. We introduce $\textbf{V}$alue-$\textbf{I}$mplicit $\textbf{P}$re-training (VIP), a self-supervised pre-trained visual representation capable of generating dense and smooth reward functions for unseen robotic tasks. VIP casts representation learning from human videos as an offline goal-conditioned reinforcement learning problem and derives a self-supervised dual goal-conditioned value-function objective that does not depend on actions, enabling pre-training on unlabeled human videos. Theoretically, VIP can be understood as a novel implicit time contrastive objective that generates a temporally smooth embedding, enabling the value function to be implicitly defined via the embedding distance, which can then be used to construct the reward for any goal-image specified downstream task. Trained on large-scale Ego4D human videos and without any fine-tuning on in-domain, task-specific data, VIP's frozen representation can provide dense visual reward for an extensive set of simulated and $\textbf{real-robot}$ tasks, enabling diverse reward-based visual control methods and significantly outperforming all prior pre-trained representations. Notably, VIP can enable simple, $\textbf{few-shot}$ offline RL on a suite of real-world robot tasks with as few as 20 trajectories.
comment: ICLR 2023, Notable-Top-25% (Spotlight). Project website: https://sites.google.com/view/vip-rl
♻ ☆ Bootstrapping Semi-supervised Medical Image Segmentation with Anatomical-aware Contrastive Distillation
Contrastive learning has shown great promise over annotation scarcity problems in the context of medical image segmentation. Existing approaches typically assume a balanced class distribution for both labeled and unlabeled medical images. However, medical image data in reality is commonly imbalanced (i.e., multi-class label imbalance), which naturally yields blurry contours and usually incorrectly labels rare objects. Moreover, it remains unclear whether all negative samples are equally negative. In this work, we present ACTION, an Anatomical-aware ConTrastive dIstillatiON framework, for semi-supervised medical image segmentation. Specifically, we first develop an iterative contrastive distillation algorithm by softly labeling the negatives rather than binary supervision between positive and negative pairs. We also capture more semantically similar features from the randomly chosen negative set compared to the positives to enforce the diversity of the sampled data. Second, we raise a more important question: Can we really handle imbalanced samples to yield better performance? Hence, the key innovation in ACTION is to learn global semantic relationship across the entire dataset and local anatomical features among the neighbouring pixels with minimal additional memory footprint. During the training, we introduce anatomical contrast by actively sampling a sparse set of hard negative pixels, which can generate smoother segmentation boundaries and more accurate predictions. Extensive experiments across two benchmark datasets and different unlabeled settings show that ACTION significantly outperforms the current state-of-the-art semi-supervised methods.
♻ ☆ View Synthesis with Sculpted Neural Points
We address the task of view synthesis, generating novel views of a scene given a set of images as input. In many recent works such as NeRF (Mildenhall et al., 2020), the scene geometry is parameterized using neural implicit representations (i.e., MLPs). Implicit neural representations have achieved impressive visual quality but have drawbacks in computational efficiency. In this work, we propose a new approach that performs view synthesis using point clouds. It is the first point-based method that achieves better visual quality than NeRF while being 100x faster in rendering speed. Our approach builds on existing works on differentiable point-based rendering but introduces a novel technique we call "Sculpted Neural Points (SNP)", which significantly improves the robustness to errors and holes in the reconstructed point cloud. We further propose to use view-dependent point features based on spherical harmonics to capture non-Lambertian surfaces, and new designs in the point-based rendering pipeline that further boost the performance. Finally, we show that our system supports fine-grained scene editing. Code is available at https://github.com/princeton-vl/SNP.
♻ ☆ Reparameterization through Spatial Gradient Scaling ICLR 2023
Reparameterization aims to improve the generalization of deep neural networks by transforming convolutional layers into equivalent multi-branched structures during training. However, there exists a gap in understanding how reparameterization may change and benefit the learning process of neural networks. In this paper, we present a novel spatial gradient scaling method to redistribute learning focus among weights in convolutional networks. We prove that spatial gradient scaling achieves the same learning dynamics as a branched reparameterization yet without introducing structural changes into the network. We further propose an analytical approach that dynamically learns scalings for each convolutional layer based on the spatial characteristics of its input feature map gauged by mutual information. Experiments on CIFAR-10, CIFAR-100, and ImageNet show that without searching for reparameterized structures, our proposed scaling method outperforms the state-of-the-art reparameterization strategies at a lower computational cost.
comment: Published at ICLR 2023. Code available at https://github.com/Ascend-Research/Reparameterization
♻ ☆ Human-in-the-Loop Mixup
Aligning model representations to humans has been found to improve robustness and generalization. However, such methods often focus on standard observational data. Synthetic data is proliferating and powering many advances in machine learning; yet, it is not always clear whether synthetic labels are perceptually aligned to humans -- rendering it likely model representations are not human aligned. We focus on the synthetic data used in mixup: a powerful regularizer shown to improve model robustness, generalization, and calibration. We design a comprehensive series of elicitation interfaces, which we release as HILL MixE Suite, and recruit 159 participants to provide perceptual judgments along with their uncertainties, over mixup examples. We find that human perceptions do not consistently align with the labels traditionally used for synthetic points, and begin to demonstrate the applicability of these findings to potentially increase the reliability of downstream models, particularly when incorporating human uncertainty. We release all elicited judgments in a new data hub we call H-Mix.
♻ ☆ Py-Feat: Python Facial Expression Analysis Toolbox
Studying facial expressions is a notoriously difficult endeavor. Recent advances in the field of affective computing have yielded impressive progress in automatically detecting facial expressions from pictures and videos. However, much of this work has yet to be widely disseminated in social science domains such as psychology. Current state of the art models require considerable domain expertise that is not traditionally incorporated into social science training programs. Furthermore, there is a notable absence of user-friendly and open-source software that provides a comprehensive set of tools and functions that support facial expression research. In this paper, we introduce Py-Feat, an open-source Python toolbox that provides support for detecting, preprocessing, analyzing, and visualizing facial expression data. Py-Feat makes it easy for domain experts to disseminate and benchmark computer vision models and also for end users to quickly process, analyze, and visualize face expression data. We hope this platform will facilitate increased use of facial expression data in human behavior research.
♻ ☆ Neural Collapse with Normalized Features: A Geometric Analysis over the Riemannian Manifold NeurIPS'22
When training overparameterized deep networks for classification tasks, it has been widely observed that the learned features exhibit a so-called "neural collapse" phenomenon. More specifically, for the output features of the penultimate layer, for each class the within-class features converge to their means, and the means of different classes exhibit a certain tight frame structure, which is also aligned with the last layer's classifier. As feature normalization in the last layer becomes a common practice in modern representation learning, in this work we theoretically justify the neural collapse phenomenon for normalized features. Based on an unconstrained feature model, we simplify the empirical loss function in a multi-class classification task into a nonconvex optimization problem over the Riemannian manifold by constraining all features and classifiers over the sphere. In this context, we analyze the nonconvex landscape of the Riemannian optimization problem over the product of spheres, showing a benign global landscape in the sense that the only global minimizers are the neural collapse solutions while all other critical points are strict saddles with negative curvature. Experimental results on practical deep networks corroborate our theory and demonstrate that better representations can be learned faster via feature normalization.
comment: The first two authors contributed to this work equally; 38 pages, 13 figures. Accepted at NeurIPS'22
♻ ☆ A Comprehensive Review of Data-Driven Co-Speech Gesture Generation
Gestures that accompany speech are an essential part of natural and efficient embodied human communication. The automatic generation of such co-speech gestures is a long-standing problem in computer animation and is considered an enabling technology in film, games, virtual social spaces, and for interaction with social robots. The problem is made challenging by the idiosyncratic and non-periodic nature of human co-speech gesture motion, and by the great diversity of communicative functions that gestures encompass. Gesture generation has seen surging interest recently, owing to the emergence of more and larger datasets of human gesture motion, combined with strides in deep-learning-based generative models, that benefit from the growing availability of data. This review article summarizes co-speech gesture generation research, with a particular focus on deep generative models. First, we articulate the theory describing human gesticulation and how it complements speech. Next, we briefly discuss rule-based and classical statistical gesture synthesis, before delving into deep learning approaches. We employ the choice of input modalities as an organizing principle, examining systems that generate gestures from audio, text, and non-linguistic input. We also chronicle the evolution of the related training data sets in terms of size, diversity, motion quality, and collection method. Finally, we identify key research challenges in gesture generation, including data availability and quality; producing human-like motion; grounding the gesture in the co-occurring speech in interaction with other speakers, and in the environment; performing gesture evaluation; and integration of gesture synthesis into applications. We highlight recent approaches to tackling the various key challenges, as well as the limitations of these approaches, and point toward areas of future development.
comment: Accepted for EUROGRAPHICS 2023
Information Retrieval
☆ Clustering large 3D volumes: A sampling-based approach
In many applications of X-ray computed tomography, an unsupervised segmentation of the reconstructed 3D volumes forms an important step in the image processing chain for further investigation of the digitized object. Therefore, the goal is to train a clustering algorithm on the volume, which produces a voxelwise classification by assigning a cluster index to each voxel. However, clustering methods, e.g., K-Means, typically have an asymptotic polynomial runtime with respect to the dataset size, and thus, these techniques are rarely applicable to large volumes. In this work, we introduce a novel clustering technique based on random sampling, which allows for the voxelwise classification of arbitrarily large volumes. The presented method conducts efficient linear passes over the data to extract a representative random sample of a fixed size on which the classifier can be trained. Then, a final linear pass performs the segmentation and assigns a cluster index to each individual voxel. Quantitative and qualitative evaluations show that excellent results can be achieved even with a very small sample size. Consequently, the unsupervised segmentation by means of clustering becomes feasible for arbitrarily large volumes.
comment: 12 pages, 8 figures
☆ A Privacy Preserving System for Movie Recommendations using Federated Learning
Recommender systems have become ubiquitous in the past years. They solve the tyranny of choice problem faced by many users, and are employed by many online businesses to drive engagement and sales. Besides other criticisms, like creating filter bubbles within social networks, recommender systems are often reproved for collecting considerable amounts of personal data. However, to personalize recommendations, personal information is fundamentally required. A recent distributed learning scheme called federated learning has made it possible to learn from personal user data without its central collection. Accordingly, we present a complete recommender system for movie recommendations, which provides privacy and thus trustworthiness on two levels: First, it is trained using federated learning and thus is, by its very nature, privacy-preserving, while still enabling individual users to benefit from global insights. And second, a novel federated learning scheme, FedQ, is employed, which not only addresses the problem of non-i.i.d. and small local datasets, but also prevents input data reconstruction attacks by aggregating client models early. To reduce the communication overhead, compression is applied, which significantly reduces the exchanged neural network updates to a fraction of their original data. We conjecture that it may also improve data privacy through its lossy quantization stage.
comment: Submitted to the ACM TORS Special Issue on Trustworthy Recommender Systems
☆ Automatically Summarizing Evidence from Clinical Trials: A Prototype Highlighting Current Challenges
We present TrialsSummarizer, a system that aims to automatically summarize evidence presented in the set of randomized controlled trials most relevant to a given query. Building on prior work, the system retrieves trial publications matching a query specifying a combination of condition, intervention(s), and outcome(s), and ranks these according to sample size and estimated study quality. The top-k such studies are passed through a neural multi-document summarization system, yielding a synopsis of these trials. We consider two architectures: A standard sequence-to-sequence model based on BART and a multi-headed architecture intended to provide greater transparency to end-users. Both models produce fluent and relevant summaries of evidence retrieved for queries, but their tendency to introduce unsupported statements render them inappropriate for use in this domain at present. The proposed architecture may help users verify outputs allowing users to trace generated tokens back to inputs.
♻ ☆ Unifying Remote Sensing Image Retrieval and Classification with Robust Fine-tuning
Advances in high resolution remote sensing image analysis are currently hampered by the difficulty of gathering enough annotated data for training deep learning methods, giving rise to a variety of small datasets and associated dataset-specific methods. Moreover, typical tasks such as classification and retrieval lack a systematic evaluation on standard benchmarks and training datasets, which make it hard to identify durable and generalizable scientific contributions. We aim at unifying remote sensing image retrieval and classification with a new large-scale training and testing dataset, SF300, including both vertical and oblique aerial images and made available to the research community, and an associated fine-tuning method. We additionally propose a new adversarial fine-tuning method for global descriptors. We show that our framework systematically achieves a boost of retrieval and classification performance on nine different datasets compared to an ImageNet pretrained baseline, with currently no other method to compare to.
comment: Performance margin with the proposed method is not statistically significant. Please refer to http://alegoria.ign.fr/en/SF300_dataset if you are interested in the dataset
♻ ☆ LambdaKG: A Library for Pre-trained Language Model-Based Knowledge Graph Embeddings
Knowledge Graphs (KGs) often have two characteristics: heterogeneous graph structure and text-rich entity/relation information. Text-based KG embeddings can represent entities by encoding descriptions with pre-trained language models, but no open-sourced library is specifically designed for KGs with PLMs at present. In this paper, we present LambdaKG, a library for KGE that equips with many pre-trained language models (e.g., BERT, BART, T5, GPT-3), and supports various tasks (e.g., knowledge graph completion, question answering, recommendation, and knowledge probing). LambdaKG is publicly open-sourced at https://github.com/zjunlp/PromptKG/tree/main/lambdaKG, with a demo video at http://deepke.zjukg.cn/lambdakg.mp4 and long-term maintenance.
comment: Work in progress and the project website is https://zjunlp.github.io/project/promptkg/
♻ ☆ MAMDR: A Model Agnostic Learning Method for Multi-Domain Recommendation ICDE 2023
Large-scale e-commercial platforms in the real-world usually contain various recommendation scenarios (domains) to meet demands of diverse customer groups. Multi-Domain Recommendation (MDR), which aims to jointly improve recommendations on all domains and easily scales to thousands of domains, has attracted increasing attention from practitioners and researchers. Existing MDR methods usually employ a shared structure and several specific components to respectively leverage reusable features and domain-specific information. However, data distribution differs across domains, making it challenging to develop a general model that can be applied to all circumstances. Additionally, during training, shared parameters often suffer from the domain conflict while specific parameters are inclined to overfitting on data sparsity domains. we first present a scalable MDR platform served in Taobao that enables to provide services for thousands of domains without specialists involved. To address the problems of MDR methods, we propose a novel model agnostic learning framework, namely MAMDR, for the multi-domain recommendation. Specifically, we first propose a Domain Negotiation (DN) strategy to alleviate the conflict between domains. Then, we develop a Domain Regularization (DR) to improve the generalizability of specific parameters by learning from other domains. We integrate these components into a unified framework and present MAMDR, which can be applied to any model structure to perform multi-domain recommendation. Finally, we present a large-scale implementation of MAMDR in the Taobao application and construct various public MDR benchmark datasets which can be used for following studies. Extensive experiments on both benchmark datasets and industry datasets demonstrate the effectiveness and generalizability of MAMDR.
comment: This paper has been accepted by ICDE 2023
♻ ☆ AugTriever: Unsupervised Dense Retrieval by Scalable Data Augmentation
Dense retrievers have made significant strides in text retrieval and open-domain question answering, even though most achievements were made possible only with large amounts of human supervision. In this work, we aim to develop unsupervised methods by proposing two methods that create pseudo query-document pairs and train dense retrieval models in an annotation-free and scalable manner: query extraction and transferred query generation. The former method produces pseudo queries by selecting salient spans from the original document. The latter utilizes generation models trained for other NLP tasks (e.g., summarization) to produce pseudo queries. Extensive experiments show that models trained with the proposed augmentation methods can perform comparably well (or better) to multiple strong baselines. Combining those strategies leads to further improvements, achieving the state-of-the-art performance of unsupervised dense retrieval on both BEIR and ODQA datasets.
Machine Learning
☆ Benign Overfitting for Two-layer ReLU Networks
Modern deep learning models with great expressive power can be trained to overfit the training data but still generalize well. This phenomenon is referred to as benign overfitting. Recently, a few studies have attempted to theoretically understand benign overfitting in neural networks. However, these works are either limited to neural networks with smooth activation functions or to the neural tangent kernel regime. How and when benign overfitting can occur in ReLU neural networks remains an open problem. In this work, we seek to answer this question by establishing algorithm-dependent risk bounds for learning two-layer ReLU convolutional neural networks with label-flipping noise. We show that, under mild conditions, the neural network trained by gradient descent can achieve near-zero training loss and Bayes optimal test risk. Our result also reveals a sharp transition between benign and harmful overfitting under different conditions on data distribution in terms of test risk. Experiments on synthetic data back up our theory.
comment: 54 pages, 2 figures, 2 tables
☆ Can We Scale Transformers to Predict Parameters of Diverse ImageNet Models?
Pretraining a neural network on a large dataset is becoming a cornerstone in machine learning that is within the reach of only a few communities with large-resources. We aim at an ambitious goal of democratizing pretraining. Towards that goal, we train and release a single neural network that can predict high quality ImageNet parameters of other neural networks. By using predicted parameters for initialization we are able to boost training of diverse ImageNet models available in PyTorch. When transferred to other datasets, models initialized with predicted parameters also converge faster and reach competitive final performance.
comment: Code and models are available at https://github.com/SamsungSAILMontreal/ghn3
☆ From Copilot to Pilot: Towards AI Supported Software Development
AI-supported programming has arrived, as shown by the introduction and successes of large language models for code, such as Copilot/Codex (Github/OpenAI) and AlphaCode (DeepMind). Above human average performance on programming challenges is now possible. However, software engineering is much more than solving programming contests. Moving beyond code completion to AI-supported software engineering will require an AI system that can, among other things, understand how to avoid code smells, to follow language idioms, and eventually (maybe!) propose rational software designs. In this study, we explore the current limitations of AI-supported code completion tools like Copilot and offer a simple taxonomy for understanding the classification of AI-supported code completion tools in this space. We first perform an exploratory study on Copilot's code suggestions for language idioms and code smells. Copilot does not follow language idioms and avoid code smells in most of our test scenarios. We then conduct additional investigation to determine the current boundaries of AI-supported code completion tools like Copilot by introducing a taxonomy of software abstraction hierarchies where 'basic programming functionality' such as code compilation and syntax checking is at the least abstract level, software architecture analysis and design are at the most abstract level. We conclude by providing a discussion on challenges for future development of AI-supported code completion tools to reach the design level of abstraction in our taxonomy.
☆ Domain Randomization for Robust, Affordable and Effective Closed-loop Control of Soft Robots
Soft robots are becoming extremely popular thanks to their intrinsic safety to contacts and adaptability. However, the potentially infinite number of Degrees of Freedom makes their modeling a daunting task, and in many cases only an approximated description is available. This challenge makes reinforcement learning (RL) based approaches inefficient when deployed on a realistic scenario, due to the large domain gap between models and the real platform. In this work, we demonstrate, for the first time, how Domain Randomization (DR) can solve this problem by enhancing RL policies with: i) a higher robustness w.r.t. environmental changes; ii) a higher affordability of learned policies when the target model differs significantly from the training model; iii) a higher effectiveness of the policy, which can even autonomously learn to exploit the environment to increase the robot capabilities (environmental constraints exploitation). Moreover, we introduce a novel algorithmic extension of previous adaptive domain randomization methods for the automatic inference of dynamics parameters for deformable objects. We provide results on four different tasks and two soft robot designs, opening interesting perspectives for future research on Reinforcement Learning for closed-loop soft robot control.
comment: Project website at https://andreaprotopapa.github.io/dr-soro/
☆ Exploiting Asymmetry for Synthetic Training Data Generation: SynthIE and the Case of Information Extraction
Large language models (LLMs) show great potential for synthetic data generation. This work shows that useful data can be synthetically generated even for tasks that cannot be solved directly by the LLM: we show that, for problems with structured outputs, it is possible to prompt an LLM to perform the task in the opposite direction, to generate plausible text for the target structure. Leveraging the asymmetry in task difficulty makes it possible to produce large-scale, high-quality data for complex tasks. We demonstrate the effectiveness of this approach on closed information extraction, where collecting ground-truth data is challenging, and no satisfactory dataset exists to date. We synthetically generate a dataset of 1.8M data points, demonstrate its superior quality compared to existing datasets in a human evaluation and use it to finetune small models (220M and 770M parameters). The models we introduce, SynthIE, outperform existing baselines of comparable size with a substantial gap of 57 and 79 absolute points in micro and macro F1, respectively. Code, data, and models are available at https://github.com/epfl-dlab/SynthIE.
☆ Foundation Models for Decision Making: Problems, Methods, and Opportunities
Foundation models pretrained on diverse data at scale have demonstrated extraordinary capabilities in a wide range of vision and language tasks. When such models are deployed in real world environments, they inevitably interface with other entities and agents. For example, language models are often used to interact with human beings through dialogue, and visual perception models are used to autonomously navigate neighborhood streets. In response to these developments, new paradigms are emerging for training foundation models to interact with other agents and perform long-term reasoning. These paradigms leverage the existence of ever-larger datasets curated for multimodal, multitask, and generalist interaction. Research at the intersection of foundation models and decision making holds tremendous promise for creating powerful new systems that can interact effectively across a diverse range of applications such as dialogue, autonomous driving, healthcare, education, and robotics. In this manuscript, we examine the scope of foundation models for decision making, and provide conceptual tools and technical background for understanding the problem space and exploring new research directions. We review recent approaches that ground foundation models in practical decision making applications through a variety of methods such as prompting, conditional generative modeling, planning, optimal control, and reinforcement learning, and discuss common challenges and open problems in the field.
☆ Wigner kernels: body-ordered equivariant machine learning without a basis
Machine-learning models based on a point-cloud representation of a physical object are ubiquitous in scientific applications and particularly well-suited to the atomic-scale description of molecules and materials. Among the many different approaches that have been pursued, the description of local atomic environments in terms of their neighbor densities has been used widely and very succesfully. We propose a novel density-based method which involves computing ``Wigner kernels''. These are fully equivariant and body-ordered kernels that can be computed iteratively with a cost that is independent of the radial-chemical basis and grows only linearly with the maximum body-order considered. This is in marked contrast to feature-space models, which comprise an exponentially-growing number of terms with increasing order of correlations. We present several examples of the accuracy of models based on Wigner kernels in chemical applications, for both scalar and tensorial targets, reaching state-of-the-art accuracy on the popular QM9 benchmark dataset, and we discuss the broader relevance of these ideas to equivariant geometric machine-learning.
☆ A Multiplicative Value Function for Safe and Efficient Reinforcement Learning
An emerging field of sequential decision problems is safe Reinforcement Learning (RL), where the objective is to maximize the reward while obeying safety constraints. Being able to handle constraints is essential for deploying RL agents in real-world environments, where constraint violations can harm the agent and the environment. To this end, we propose a safe model-free RL algorithm with a novel multiplicative value function consisting of a safety critic and a reward critic. The safety critic predicts the probability of constraint violation and discounts the reward critic that only estimates constraint-free returns. By splitting responsibilities, we facilitate the learning task leading to increased sample efficiency. We integrate our approach into two popular RL algorithms, Proximal Policy Optimization and Soft Actor-Critic, and evaluate our method in four safety-focused environments, including classical RL benchmarks augmented with safety constraints and robot navigation tasks with images and raw Lidar scans as observations. Finally, we make the zero-shot sim-to-real transfer where a differential drive robot has to navigate through a cluttered room. Our code can be found at https://github.com/nikeke19/Safe-Mult-RL.
comment: Repository available at https://github.com/nikeke19/Safe-Mult-RL
☆ Validation of a Hospital Digital Twin with Machine Learning
Recently there has been a surge of interest in developing Digital Twins of process flows in healthcare to better understand bottlenecks and areas of improvement. A key challenge is in the validation process. We describe a work in progress for a digital twin using an agent based simulation model for determining bed turnaround time for patients in hospitals. We employ a strategy using machine learning for validating the model and implementing sensitivity analysis.
☆ Predicted Embedding Power Regression for Large-Scale Out-of-Distribution Detection
Out-of-distribution (OOD) inputs can compromise the performance and safety of real world machine learning systems. While many methods exist for OOD detection and work well on small scale datasets with lower resolution and few classes, few methods have been developed for large-scale OOD detection. Existing large-scale methods generally depend on maximum classification probability, such as the state-of-the-art grouped softmax method. In this work, we develop a novel approach that calculates the probability of the predicted class label based on label distributions learned during the training process. Our method performs better than current state-of-the-art methods with only a negligible increase in compute cost. We evaluate our method against contemporary methods across $14$ datasets and achieve a statistically significant improvement with respect to AUROC (84.2 vs 82.4) and AUPR (96.2 vs 93.7).
☆ Introspective Cross-Attention Probing for Lightweight Transfer of Pre-trained Models
We propose InCA, a lightweight method for transfer learning that cross-attends to any activation layer of a pre-trained model. During training, InCA uses a single forward pass to extract multiple activations, which are passed to external cross-attention adapters, trained anew and combined or selected for downstream tasks. We show that, even when selecting a single top-scoring adapter, InCA achieves performance comparable to full fine-tuning, at a cost comparable to fine-tuning just the last layer. For example, with a cross-attention probe 1.3% the size of a pre-trained ViT-L/16 model, we achieve performance within 0.2% of the full fine-tuning paragon at 51% training cost of the baseline, on average across 11 downstream classification tasks. Unlike other forms of efficient adaptation, InCA does not require backpropagating through the pre-trained model, thus leaving its execution unaltered at both training and inference. The versatility of InCA is best illustrated in fine-grained tasks, which may require accessing information absent in the last layer but accessible in intermediate layer activations. Since the backbone is fixed, InCA allows parallel ensembling as well as parallel execution of multiple tasks. InCA achieves state-of-the-art performance in the ImageNet-to-Sketch multi-task benchmark.
☆ An Inception-Residual-Based Architecture with Multi-Objective Loss for Detecting Respiratory Anomalies
This paper presents a deep learning system applied for detecting anomalies from respiratory sound recordings. Initially, our system begins with audio feature extraction using Gammatone and Continuous Wavelet transformation. This step aims to transform the respiratory sound input into a two-dimensional spectrogram where both spectral and temporal features are presented. Then, our proposed system integrates Inception-residual-based backbone models combined with multi-head attention and multi-objective loss to classify respiratory anomalies. In this work, we conducted experiments over the benchmark dataset of SPRSound (The Open-Source SJTU Paediatric Respiratory Sound) proposed by the IEEE BioCAS 2022 challenge. As regards the Score computed by an average between the average score and harmonic score, our proposed system gained significant improvements of 9.7%, 15.8%, 17.0%, and 9.4% in Task 1-1, Task 1-2, Task 2-1, and Task 2-2 compared to the challenge baseline system. Notably, we achieved the Top-1 performance in Task 2-1 with the highest Score of 73.7%.
☆ Mastering Strategy Card Game (Legends of Code and Magic) via End-to-End Policy and Optimistic Smooth Fictitious Play
Deep Reinforcement Learning combined with Fictitious Play shows impressive results on many benchmark games, most of which are, however, single-stage. In contrast, real-world decision making problems may consist of multiple stages, where the observation spaces and the action spaces can be completely different across stages. We study a two-stage strategy card game Legends of Code and Magic and propose an end-to-end policy to address the difficulties that arise in multi-stage game. We also propose an optimistic smooth fictitious play algorithm to find the Nash Equilibrium for the two-player game. Our approach wins double championships of COG2022 competition. Extensive studies verify and show the advancement of our approach.
☆ Visual Abstraction and Reasoning through Language
While Artificial Intelligence (AI) models have achieved human or even superhuman performance in narrowly defined applications, they still struggle to show signs of broader and more flexible intelligence. The Abstraction and Reasoning Corpus (ARC), introduced by Fran\c{c}ois Chollet, aims to assess how close AI systems are to human-like cognitive abilities. Most current approaches rely on carefully handcrafted domain-specific languages (DSLs), which are used to brute-force solutions to the tasks present in ARC. In this work, we propose a general framework for solving ARC based on natural language descriptions of the tasks. While not yet beating state-of-the-art DSL models on ARC, we demonstrate the immense potential of our approach hinted at by the ability to solve previously unsolved tasks.
comment: The first two authors have contributed equally to this work
☆ Uncertainty Quantification of Spatiotemporal Travel Demand with Probabilistic Graph Neural Networks
Recent studies have significantly improved the prediction accuracy of travel demand using graph neural networks. However, these studies largely ignored uncertainty that inevitably exists in travel demand prediction. To fill this gap, this study proposes a framework of probabilistic graph neural networks (Prob-GNN) to quantify the spatiotemporal uncertainty of travel demand. This Prob-GNN framework is substantiated by deterministic and probabilistic assumptions, and empirically applied to the task of predicting the transit and ridesharing demand in Chicago. We found that the probabilistic assumptions (e.g. distribution tail, support) have a greater impact on uncertainty prediction than the deterministic ones (e.g. deep modules, depth). Among the family of Prob-GNNs, the GNNs with truncated Gaussian and Laplace distributions achieve the highest performance in transit and ridesharing data. Even under significant domain shifts, Prob-GNNs can predict the ridership uncertainty in a stable manner, when the models are trained on pre-COVID data and tested across multiple periods during and after the COVID-19 pandemic. Prob-GNNs also reveal the spatiotemporal pattern of uncertainty, which is concentrated on the afternoon peak hours and the areas with large travel volumes. Overall, our findings highlight the importance of incorporating randomness into deep learning for spatiotemporal ridership prediction. Future research should continue to investigate versatile probabilistic assumptions to capture behavioral randomness, and further develop methods to quantify uncertainty to build resilient cities.
☆ Root Cause Identification for Collective Anomalies in Time Series given an Acyclic Summary Causal Graph with Loops AISTATS
This paper presents an approach for identifying the root causes of collective anomalies given observational time series and an acyclic summary causal graph which depicts an abstraction of causal relations present in a dynamic system at its normal regime. The paper first shows how the problem of root cause identification can be divided into many independent subproblems by grouping related anomalies using d-separation. Further, it shows how, under this setting, some root causes can be found directly from the graph and from the time of appearance of anomalies. Finally, it shows, how the rest of the root causes can be found by comparing direct causal effects in the normal and in the anomalous regime. To this end, temporal adaptations of the back-door and the single-door criterions are introduced. Extensive experiments conducted on both simulated and real-world datasets demonstrate the effectiveness of the proposed method.
comment: Proceedings of the 26th International Conference on Artificial Intelligence and Statistics (AISTATS) 2023, Valencia, Spain
☆ PyXAB -- A Python Library for $\mathcal{X}$-Armed Bandit and Online Blackbox Optimization Algorithms
We introduce a Python open-source library for $\mathcal{X}$-armed bandit and online blackbox optimization named PyXAB. PyXAB contains the implementations for more than 10 $\mathcal{X}$-armed bandit algorithms, such as HOO, StoSOO, HCT, and the most recent works GPO and VHCT. PyXAB also provides the most commonly-used synthetic objectives to evaluate the performance of different algorithms and the various choices of the hierarchical partitions on the parameter space. The online documentation for PyXAB includes clear instructions for installation, straight-forward examples, detailed feature descriptions, and a complete reference of the API. PyXAB is released under the MIT license in order to encourage both academic and industrial usage. The library can be directly installed from PyPI with its source code available at https://github.com/WilliamLwj/PyXAB
☆ When is Importance Weighting Correction Needed for Covariate Shift Adaptation?
This paper investigates when the importance weighting (IW) correction is needed to address covariate shift, a common situation in supervised learning where the input distributions of training and test data differ. Classic results show that the IW correction is needed when the model is parametric and misspecified. In contrast, recent results indicate that the IW correction may not be necessary when the model is nonparametric and well-specified. We examine the missing case in the literature where the model is nonparametric and misspecified, and show that the IW correction is needed for obtaining the best approximation of the true unknown function for the test distribution. We do this by analyzing IW-corrected kernel ridge regression, covering a variety of settings, including parametric and nonparametric models, well-specified and misspecified settings, and arbitrary weighting functions.
☆ Decoupling Skill Learning from Robotic Control for Generalizable Object Manipulation ICRA
Recent works in robotic manipulation through reinforcement learning (RL) or imitation learning (IL) have shown potential for tackling a range of tasks e.g., opening a drawer or a cupboard. However, these techniques generalize poorly to unseen objects. We conjecture that this is due to the high-dimensional action space for joint control. In this paper, we take an alternative approach and separate the task of learning 'what to do' from 'how to do it' i.e., whole-body control. We pose the RL problem as one of determining the skill dynamics for a disembodied virtual manipulator interacting with articulated objects. The whole-body robotic kinematic control is optimized to execute the high-dimensional joint motion to reach the goals in the workspace. It does so by solving a quadratic programming (QP) model with robotic singularity and kinematic constraints. Our experiments on manipulating complex articulated objects show that the proposed approach is more generalizable to unseen objects with large intra-class variations, outperforming previous approaches. The evaluation results indicate that our approach generates more compliant robotic motion and outperforms the pure RL and IL baselines in task success rates.
comment: Accepted to IEEE International Conference on Robotics and Automation (ICRA) 2023
☆ Exploration via Epistemic Value Estimation
How to efficiently explore in reinforcement learning is an open problem. Many exploration algorithms employ the epistemic uncertainty of their own value predictions -- for instance to compute an exploration bonus or upper confidence bound. Unfortunately the required uncertainty is difficult to estimate in general with function approximation. We propose epistemic value estimation (EVE): a recipe that is compatible with sequential decision making and with neural network function approximators. It equips agents with a tractable posterior over all their parameters from which epistemic value uncertainty can be computed efficiently. We use the recipe to derive an epistemic Q-Learning agent and observe competitive performance on a series of benchmarks. Experiments confirm that the EVE recipe facilitates efficient exploration in hard exploration tasks.
☆ ELODIN: Naming Concepts in Embedding Spaces
Despite recent advancements, the field of text-to-image synthesis still suffers from lack of fine-grained control. Using only text, it remains challenging to deal with issues such as concept coherence and concept contamination. We propose a method to enhance control by generating specific concepts that can be reused throughout multiple images, effectively expanding natural language with new words that can be combined much like a painter's palette. Unlike previous contributions, our method does not copy visuals from input data and can generate concepts through text alone. We perform a set of comparisons that finds our method to be a significant improvement over text-only prompts.
☆ Group conditional validity via multi-group learning
We consider the problem of distribution-free conformal prediction and the criterion of group conditional validity. This criterion is motivated by many practical scenarios including hidden stratification and group fairness. Existing methods achieve such guarantees under either restrictive grouping structure or distributional assumptions, or they are overly-conservative under heteroskedastic noise. We propose a simple reduction to the problem of achieving validity guarantees for individual populations by leveraging algorithms for a problem called multi-group learning. This allows us to port theoretical guarantees from multi-group learning to obtain obtain sample complexity guarantees for conformal prediction. We also provide a new algorithm for multi-group learning for groups with hierarchical structure. Using this algorithm in our reduction leads to improved sample complexity guarantees with a simpler predictor structure.
☆ Enhanced Adaptive Gradient Algorithms for Nonconvex-PL Minimax Optimization
In the paper, we study a class of nonconvex nonconcave minimax optimization problems (i.e., $\min_x\max_y f(x,y)$), where $f(x,y)$ is possible nonconvex in $x$, and it is nonconcave and satisfies the Polyak-Lojasiewicz (PL) condition in $y$. Moreover, we propose a class of enhanced momentum-based gradient descent ascent methods (i.e., MSGDA and AdaMSGDA) to solve these stochastic Nonconvex-PL minimax problems. In particular, our AdaMSGDA algorithm can use various adaptive learning rates in updating the variables $x$ and $y$ without relying on any global and coordinate-wise adaptive learning rates. Theoretically, we present an effective convergence analysis framework for our methods. Specifically, we prove that our MSGDA and AdaMSGDA methods have the best known sample (gradient) complexity of $O(\epsilon^{-3})$ only requiring one sample at each loop in finding an $\epsilon$-stationary solution (i.e., $\mathbb{E}\|\nabla F(x)\|\leq \epsilon$, where $F(x)=\max_y f(x,y)$). This manuscript commemorates the mathematician Boris Polyak (1935-2023).
comment: 30 pages
☆ Structured State Space Models for In-Context Reinforcement Learning
Structured state space sequence (S4) models have recently achieved state-of-the-art performance on long-range sequence modeling tasks. These models also have fast inference speeds and parallelisable training, making them potentially useful in many reinforcement learning settings. We propose a modification to a variant of S4 that enables us to initialise and reset the hidden state in parallel, allowing us to tackle reinforcement learning tasks. We show that our modified architecture runs asymptotically faster than Transformers and performs better than LSTM models on a simple memory-based task. Then, by leveraging the model's ability to handle long-range sequences, we achieve strong performance on a challenging meta-learning task in which the agent is given a randomly-sampled continuous control environment, combined with a randomly-sampled linear projection of the environment's observations and actions. Furthermore, we show the resulting model can adapt to out-of-distribution held-out tasks. Overall, the results presented in this paper suggest that the S4 models are a strong contender for the default architecture used for in-context reinforcement learning
☆ Comparing 3D deformations between longitudinal daily CBCT acquisitions using CNN for head and neck radiotherapy toxicity prediction
Adaptive radiotherapy is a growing field of study in cancer treatment due to it's objective in sparing healthy tissue. The standard of care in several institutions includes longitudinal cone-beam computed tomography (CBCT) acquisitions to monitor changes, but have yet to be used to improve tumor control while managing side-effects. The aim of this study is to demonstrate the clinical value of pre-treatment CBCT acquired daily during radiation therapy treatment for head and neck cancers for the downstream task of predicting severe toxicity occurrence: reactive feeding tube (NG), hospitalization and radionecrosis. For this, we propose a deformable 3D classification pipeline that includes a component analyzing the Jacobian matrix of the deformation between planning CT and longitudinal CBCT, as well as clinical data. The model is based on a multi-branch 3D residual convolutional neural network, while the CT to CBCT registration is based on a pair of VoxelMorph architectures. Accuracies of 85.8% and 75.3% was found for radionecrosis and hospitalization, respectively, with similar performance as early as after the first week of treatment. For NG tube risk, performance improves with increasing the timing of the CBCT fraction, reaching 83.1% after the $5_{th}$ week of treatment.
comment: 11 pages, 3 figures, 2 equations, 2 tables
☆ Diminishing Return of Value Expansion Methods in Model-Based Reinforcement Learning ICLR 2023
Model-based reinforcement learning is one approach to increase sample efficiency. However, the accuracy of the dynamics model and the resulting compounding error over modelled trajectories are commonly regarded as key limitations. A natural question to ask is: How much more sample efficiency can be gained by improving the learned dynamics models? Our paper empirically answers this question for the class of model-based value expansion methods in continuous control problems. Value expansion methods should benefit from increased model accuracy by enabling longer rollout horizons and better value function approximations. Our empirical study, which leverages oracle dynamics models to avoid compounding model errors, shows that (1) longer horizons increase sample efficiency, but the gain in improvement decreases with each additional expansion step, and (2) the increased model accuracy only marginally increases the sample efficiency compared to learned models with identical horizons. Therefore, longer horizons and increased model accuracy yield diminishing returns in terms of sample efficiency. These improvements in sample efficiency are particularly disappointing when compared to model-free value expansion methods. Even though they introduce no computational overhead, we find their performance to be on-par with model-based value expansion methods. Therefore, we conclude that the limitation of model-based value expansion methods is not the model accuracy of the learned models. While higher model accuracy is beneficial, our experiments show that even a perfect model will not provide an un-rivalled sample efficiency but that the bottleneck lies elsewhere.
comment: Published as a conference paper at ICLR 2023
☆ ChatGPT: Beginning of an End of Manual Annotation? Use Case of Automatic Genre Identification
ChatGPT has shown strong capabilities in natural language generation tasks, which naturally leads researchers to explore where its abilities end. In this paper, we examine whether ChatGPT can be used for zero-shot text classification, more specifically, automatic genre identification. We compare ChatGPT with a multilingual XLM-RoBERTa language model that was fine-tuned on datasets, manually annotated with genres. The models are compared on test sets in two languages: English and Slovenian. Results show that ChatGPT outperforms the fine-tuned model when applied to the dataset which was not seen before by either of the models. Even when applied on Slovenian language as an under-resourced language, ChatGPT's performance is no worse than when applied to English. However, if the model is fully prompted in Slovenian, the performance drops significantly, showing the current limitations of ChatGPT usage on smaller languages. The presented results lead us to questioning whether this is the beginning of an end of laborious manual annotation campaigns even for smaller languages, such as Slovenian.
☆ Probing Graph Representations AISTATS 2023
Today we have a good theoretical understanding of the representational power of Graph Neural Networks (GNNs). For example, their limitations have been characterized in relation to a hierarchy of Weisfeiler-Lehman (WL) isomorphism tests. However, we do not know what is encoded in the learned representations. This is our main question. We answer it using a probing framework to quantify the amount of meaningful information captured in graph representations. Our findings on molecular datasets show the potential of probing for understanding the inductive biases of graph-based models. We compare different families of models and show that transformer-based models capture more chemically relevant information compared to models based on message passing. We also study the effect of different design choices such as skip connections and virtual nodes. We advocate for probing as a useful diagnostic tool for evaluating graph-based models.
comment: 20 pages, 12 figures, AISTATS 2023
☆ On Momentum-Based Gradient Methods for Bilevel Optimization with Nonconvex Lower-Level
Bilevel optimization is a popular two-level hierarchical optimization, which has been widely applied to many machine learning tasks such as hyperparameter learning, meta learning and continual learning. Although many bilevel optimization methods recently have been developed, the bilevel methods are not well studied when the lower-level problem is nonconvex. To fill this gap, in the paper, we study a class of nonconvex bilevel optimization problems, which both upper-level and lower-level problems are nonconvex, and the lower-level problem satisfies Polyak-Lojasiewicz (PL) condition. We propose an efficient momentum-based gradient bilevel method (MGBiO) to solve these deterministic problems. Meanwhile, we propose a class of efficient momentum-based stochastic gradient bilevel methods (MSGBiO and VR-MSGBiO) to solve these stochastic problems. Moreover, we provide a useful convergence analysis framework for our methods. Specifically, under some mild conditions, we prove that our MGBiO method has a sample (or gradient) complexity of $O(\epsilon^{-2})$ for finding an $\epsilon$-stationary solution of the deterministic bilevel problems (i.e., $\|\nabla F(x)\|\leq \epsilon$), which improves the existing best results by a factor of $O(\epsilon^{-1})$. Meanwhile, we prove that our MSGBiO and VR-MSGBiO methods have sample complexities of $\tilde{O}(\epsilon^{-4})$ and $\tilde{O}(\epsilon^{-3})$, respectively, in finding an $\epsilon$-stationary solution of the stochastic bilevel problems (i.e., $\mathbb{E}\|\nabla F(x)\|\leq \epsilon$), which improves the existing best results by a factor of $O(\epsilon^{-3})$. This manuscript commemorates the mathematician Boris Polyak (1935 -2023).
comment: 37 pages
☆ Fast Latent Factor Analysis via a Fuzzy PID-Incorporated Stochastic Gradient Descent Algorithm
A high-dimensional and incomplete (HDI) matrix can describe the complex interactions among numerous nodes in various big data-related applications. A stochastic gradient descent (SGD)-based latent factor analysis (LFA) model is remarkably effective in extracting valuable information from an HDI matrix. However, such a model commonly encounters the problem of slow convergence because a standard SGD algorithm learns a latent factor relying on the stochastic gradient of current instance error only without considering past update information. To address this critical issue, this paper innovatively proposes a Fuzzy PID-incorporated SGD (FPS) algorithm with two-fold ideas: 1) rebuilding the instance learning error by considering the past update information in an efficient way following the principle of PID, and 2) implementing hyper-parameters and gain parameters adaptation following the fuzzy rules. With it, an FPS-incorporated LFA model is further achieved for fast processing an HDI matrix. Empirical studies on six HDI datasets demonstrate that the proposed FPS-incorporated LFA model significantly outperforms the state-of-the-art LFA models in terms of computational efficiency for predicting the missing data of an HDI matrix with competitive accuracy.
comment: 6 pages
☆ FFT-based Dynamic Token Mixer for Vision
Multi-head-self-attention (MHSA)-equipped models have achieved notable performance in computer vision. Their computational complexity is proportional to quadratic numbers of pixels in input feature maps, resulting in slow processing, especially when dealing with high-resolution images. New types of token-mixer are proposed as an alternative to MHSA to circumvent this problem: an FFT-based token-mixer, similar to MHSA in global operation but with lower computational complexity. However, despite its attractive properties, the FFT-based token-mixer has not been carefully examined in terms of its compatibility with the rapidly evolving MetaFormer architecture. Here, we propose a novel token-mixer called dynamic filter and DFFormer and CDFFormer, image recognition models using dynamic filters to close the gaps above. CDFFormer achieved a Top-1 accuracy of 85.0%, close to the hybrid architecture with convolution and MHSA. Other wide-ranging experiments and analysis, including object detection and semantic segmentation, demonstrate that they are competitive with state-of-the-art architectures; Their throughput and memory efficiency when dealing with high-resolution image recognition is convolution and MHSA, not much different from ConvFormer, and far superior to CAFormer. Our results indicate that the dynamic filter is one of the token-mixer options that should be seriously considered. The code is available at https://github.com/okojoalg/dfformer
☆ Document-level Relation Extraction with Cross-sentence Reasoning Graph PAKDD 2023
Relation extraction (RE) has recently moved from the sentence-level to document-level, which requires aggregating document information and using entities and mentions for reasoning. Existing works put entity nodes and mention nodes with similar representations in a document-level graph, whose complex edges may incur redundant information. Furthermore, existing studies only focus on entity-level reasoning paths without considering global interactions among entities cross-sentence. To these ends, we propose a novel document-level RE model with a GRaph information Aggregation and Cross-sentence Reasoning network (GRACR). Specifically, a simplified document-level graph is constructed to model the semantic information of all mentions and sentences in a document, and an entity-level graph is designed to explore relations of long-distance cross-sentence entity pairs. Experimental results show that GRACR achieves excellent performance on two public datasets of document-level RE. It is especially effective in extracting potential relations of cross-sentence entity pairs. Our code is available at https://github.com/UESTC-LHF/GRACR.
comment: This paper is accepted by PAKDD 2023
☆ Client-specific Property Inference against Secure Aggregation in Federated Learning
Federated learning has become a widely used paradigm for collaboratively training a common model among different participants with the help of a central server that coordinates the training. Although only the model parameters or other model updates are exchanged during the federated training instead of the participant's data, many attacks have shown that it is still possible to infer sensitive information such as membership, property, or outright reconstruction of participant data. Although differential privacy is considered an effective solution to protect against privacy attacks, it is also criticized for its negative effect on utility. Another possible defense is to use secure aggregation which allows the server to only access the aggregated update instead of each individual one, and it is often more appealing because it does not degrade model quality. However, combining only the aggregated updates, which are generated by a different composition of clients in every round, may still allow the inference of some client-specific information. In this paper, we show that simple linear models can effectively capture client-specific properties only from the aggregated model updates due to the linearity of aggregation. We formulate an optimization problem across different rounds in order to infer a tested property of every client from the output of the linear models, for example, whether they have a specific sample in their training data (membership inference) or whether they misbehave and attempt to degrade the performance of the common model by poisoning attacks. Our reconstruction technique is completely passive and undetectable. We demonstrate the efficacy of our approach on several scenarios which shows that secure aggregation provides very limited privacy guarantees in practice. The source code will be released upon publication.
☆ GaussianMLR: Learning Implicit Class Significance via Calibrated Multi-Label Ranking
Existing multi-label frameworks only exploit the information deduced from the bipartition of the labels into a positive and negative set. Therefore, they do not benefit from the ranking order between positive labels, which is the concept we introduce in this paper. We propose a novel multi-label ranking method: GaussianMLR, which aims to learn implicit class significance values that determine the positive label ranks instead of treating them as of equal importance, by following an approach that unifies ranking and classification tasks associated with multi-label ranking. Due to the scarcity of public datasets, we introduce eight synthetic datasets generated under varying importance factors to provide an enriched and controllable experimental environment for this study. On both real-world and synthetic datasets, we carry out extensive comparisons with relevant baselines and evaluate the performance on both of the two sub-tasks. We show that our method is able to accurately learn a representation of the incorporated positive rank order, which is not only consistent with the ground truth but also proportional to the underlying information. We strengthen our claims empirically by conducting comprehensive experimental studies. Code is available at https://github.com/MrGranddy/GaussianMLR.
☆ Manually Selecting The Data Function for Supervised Learning of small datasets
Supervised learning problems may become ill-posed when there is a lack of information, resulting in unstable and non-unique solutions. However, instead of solely relying on regularization, initializing an informative ill-posed operator is akin to posing better questions to achieve more accurate answers. The Fredholm integral equation of the first kind (FIFK) is a reliable ill-posed operator that can integrate distributions and prior knowledge as input information. By incorporating input distributions and prior knowledge, the FIFK operator can address the limitations of using high-dimensional input distributions by semi-supervised assumptions, leading to more precise approximations of the integral operator. Additionally, the FIFK's incorporation of probabilistic principles can further enhance the accuracy and effectiveness of solutions. In cases of noisy operator equations and limited data, the FIFK's flexibility in defining problems using prior information or cross-validation with various kernel designs is especially advantageous. This capability allows for detailed problem definitions and facilitates achieving high levels of accuracy and stability in solutions. In our study, we examined the FIFK through two different approaches. Firstly, we implemented a semi-supervised assumption by using the same Fredholm operator kernel and data function kernel and incorporating unlabeled information. Secondly, we used the MSDF method, which involves selecting different kernels on both sides of the equation to define when the mapping kernel is different from the data function kernel. To assess the effectiveness of the FIFK and the proposed methods in solving ill-posed problems, we conducted experiments on a real-world dataset. Our goal was to compare the performance of these methods against the widely used least-squares method and other comparable methods.
☆ Organelle-specific segmentation, spatial analysis, and visualization of volume electron microscopy datasets
Volume electron microscopy is the method of choice for the in-situ interrogation of cellular ultrastructure at the nanometer scale. Recent technical advances have led to a rapid increase in large raw image datasets that require computational strategies for segmentation and spatial analysis. In this protocol, we describe a practical and annotation-efficient pipeline for organelle-specific segmentation, spatial analysis, and visualization of large volume electron microscopy datasets using freely available, user-friendly software tools that can be run on a single standard workstation. We specifically target researchers in the life sciences with limited computational expertise, who face the following tasks within their volume electron microscopy projects: i) How to generate 3D segmentation labels for different types of cell organelles while minimizing manual annotation efforts, ii) how to analyze the spatial interactions between organelle instances, and iii) how to best visualize the 3D segmentation results. To meet these demands we give detailed guidelines for choosing the most efficient segmentation tools for the specific cell organelle. We furthermore provide easily executable components for spatial analysis and 3D rendering and bridge compatibility issues between freely available open-source tools, such that others can replicate our full pipeline starting from a raw dataset up to the final plots and rendered images. We believe that our detailed description can serve as a valuable reference for similar projects requiring special strategies for single- or multiple organelle analysis which can be achieved with computational resources commonly available to single-user setups.
☆ Can Decentralized Learning be more robust than Federated Learning?
Decentralized Learning (DL) is a peer--to--peer learning approach that allows a group of users to jointly train a machine learning model. To ensure correctness, DL should be robust, i.e., Byzantine users must not be able to tamper with the result of the collaboration. In this paper, we introduce two \textit{new} attacks against DL where a Byzantine user can: make the network converge to an arbitrary model of their choice, and exclude an arbitrary user from the learning process. We demonstrate our attacks' efficiency against Self--Centered Clipping, the state--of--the--art robust DL protocol. Finally, we show that the capabilities decentralization grants to Byzantine users result in decentralized learning \emph{always} providing less robustness than federated learning.
☆ ENTROPY: Environment Transformer and Offline Policy Optimization
Model-based methods provide an effective approach to offline reinforcement learning (RL). They learn an environmental dynamics model from interaction experiences and then perform policy optimization based on the learned model. However, previous model-based offline RL methods lack long-term prediction capability, resulting in large errors when generating multi-step trajectories. We address this issue by developing a sequence modeling architecture, Environment Transformer, which can generate reliable long-horizon trajectories based on offline datasets. We then propose a novel model-based offline RL algorithm, ENTROPY, that learns the dynamics model and reward function by ENvironment TRansformer and performs Offline PolicY optimization. We evaluate the proposed method on MuJoCo continuous control RL environments. Results show that ENTROPY performs comparably or better than the state-of-the-art model-based and model-free offline RL methods and demonstrates more powerful long-term trajectory prediction capability compared to existing model-based offline methods.
☆ Fast and Multi-aspect Mining of Complex Time-stamped Event Streams WWW 2023
Given a huge, online stream of time-evolving events with multiple attributes, such as online shopping logs: (item, price, brand, time), and local mobility activities: (pick-up and drop-off locations, time), how can we summarize large, dynamic high-order tensor streams? How can we see any hidden patterns, rules, and anomalies? Our answer is to focus on two types of patterns, i.e., ''regimes'' and ''components'', for which we present CubeScope, an efficient and effective method over high-order tensor streams. Specifically, it identifies any sudden discontinuity and recognizes distinct dynamical patterns, ''regimes'' (e.g., weekday/weekend/holiday patterns). In each regime, it also performs multi-way summarization for all attributes (e.g., item, price, brand, and time) and discovers hidden ''components'' representing latent groups (e.g., item/brand groups) and their relationship. Thanks to its concise but effective summarization, CubeScope can also detect the sudden appearance of anomalies and identify the types of anomalies that occur in practice. Our proposed method has the following properties: (a) Effective: it captures dynamical multi-aspect patterns, i.e., regimes and components, and statistically summarizes all the events; (b) General: it is practical for successful application to data compression, pattern discovery, and anomaly detection on various types of tensor streams; (c) Scalable: our algorithm does not depend on the length of the data stream and its dimensionality. Extensive experiments on real datasets demonstrate that CubeScope finds meaningful patterns and anomalies correctly, and consistently outperforms the state-of-the-art methods as regards accuracy and execution speed.
comment: Accepted by WWW 2023
☆ Sample-efficient Real-time Planning with Curiosity Cross-Entropy Method and Contrastive Learning
Model-based reinforcement learning (MBRL) with real-time planning has shown great potential in locomotion and manipulation control tasks. However, the existing planning methods, such as the Cross-Entropy Method (CEM), do not scale well to complex high-dimensional environments. One of the key reasons for underperformance is the lack of exploration, as these planning methods only aim to maximize the cumulative extrinsic reward over the planning horizon. Furthermore, planning inside the compact latent space in the absence of observations makes it challenging to use curiosity-based intrinsic motivation. We propose Curiosity CEM (CCEM), an improved version of the CEM algorithm for encouraging exploration via curiosity. Our proposed method maximizes the sum of state-action Q values over the planning horizon, in which these Q values estimate the future extrinsic and intrinsic reward, hence encouraging reaching novel observations. In addition, our model uses contrastive representation learning to efficiently learn latent representations. Experiments on image-based continuous control tasks from the DeepMind Control suite show that CCEM is by a large margin more sample-efficient than previous MBRL algorithms and compares favorably with the best model-free RL methods.
comment: 7 pages, 4 figures
☆ Guiding Pseudo-labels with Uncertainty Estimation for Test-Time Adaptation CVPR
Standard Unsupervised Domain Adaptation (UDA) methods assume the availability of both source and target data during the adaptation. In this work, we investigate the Test-Time Adaptation (TTA), a specific case of UDA where a model is adapted to a target domain without access to source data. We propose a novel approach for the TTA setting based on a loss reweighting strategy that brings robustness against the noise that inevitably affects the pseudo-labels. The classification loss is reweighted based on the reliability of the pseudo-labels that is measured by estimating their uncertainty. Guided by such reweighting strategy, the pseudo-labels are progressively refined by aggregating knowledge from neighbouring samples. Furthermore, a self-supervised contrastive framework is leveraged as a target space regulariser to enhance such knowledge aggregation. A novel negative pairs exclusion strategy is proposed to identify and exclude negative pairs made of samples sharing the same class, even in presence of some noise in the pseudo-labels. Our method outperforms previous methods on three major benchmarks by a large margin. We set the new TTA state-of-the-art on VisDA-C and DomainNet with a performance gain of +1.8\% on both benchmarks and on PACS with +12.3\% in the single-source setting and +6.6\% in\ multi-target adaptation. Additional analyses demonstrate that the proposed approach is robust to the noise, which results in significantly more accurate pseudo-labels compared to state-of-the-art approaches.
comment: To be published in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2023
☆ Learning Hamiltonian Systems with Mono-Implicit Runge-Kutta Methods
Numerical integrators could be used to form interpolation conditions when training neural networks to approximate the vector field of an ordinary differential equation (ODE) from data. When numerical one-step schemes such as the Runge-Kutta methods are used to approximate the temporal discretization of an ODE with a known vector field, properties such as symmetry and stability are much studied. Here, we show that using mono-implicit Runge-Kutta methods of high order allows for accurate training of Hamiltonian neural networks on small datasets. This is demonstrated by numerical experiments where the Hamiltonian of the chaotic double pendulum in addition to the Fermi-Pasta-Ulam-Tsingou system is learned from data.
comment: 8 pages, 4 figures
☆ Proactive Multi-Camera Collaboration For 3D Human Pose Estimation ICLR 2023
This paper presents a multi-agent reinforcement learning (MARL) scheme for proactive Multi-Camera Collaboration in 3D Human Pose Estimation in dynamic human crowds. Traditional fixed-viewpoint multi-camera solutions for human motion capture (MoCap) are limited in capture space and susceptible to dynamic occlusions. Active camera approaches proactively control camera poses to find optimal viewpoints for 3D reconstruction. However, current methods still face challenges with credit assignment and environment dynamics. To address these issues, our proposed method introduces a novel Collaborative Triangulation Contribution Reward (CTCR) that improves convergence and alleviates multi-agent credit assignment issues resulting from using 3D reconstruction accuracy as the shared reward. Additionally, we jointly train our model with multiple world dynamics learning tasks to better capture environment dynamics and encourage anticipatory behaviors for occlusion avoidance. We evaluate our proposed method in four photo-realistic UE4 environments to ensure validity and generalizability. Empirical results show that our method outperforms fixed and active baselines in various scenarios with different numbers of cameras and humans.
comment: ICLR 2023 poster
☆ Graph Neural Networks in Vision-Language Image Understanding: A Survey
2D image understanding is a complex problem within Computer Vision, but it holds the key to providing human level scene comprehension. It goes further than identifying the objects in an image, and instead it attempts to understand the scene. Solutions to this problem form the underpinning of a range of tasks, including image captioning, Visual Question Answering (VQA), and image retrieval. Graphs provide a natural way to represent the relational arrangement between objects in an image, and thus in recent years Graph Neural Networks (GNNs) have become a standard component of many 2D image understanding pipelines, becoming a core architectural component especially in the VQA group of tasks. In this survey, we review this rapidly evolving field and we provide a taxonomy of graph types used in 2D image understanding approaches, a comprehensive list of the GNN models used in this domain, and a roadmap of future potential developments. To the best of our knowledge, this is the first comprehensive survey that covers image captioning, visual question answering, and image retrieval techniques that focus on using GNNs as the main part of their architecture.
comment: 19 pages, 5 figures, 6 tables
☆ DLT: Conditioned layout generation with Joint Discrete-Continuous Diffusion Layout Transformer
Generating visual layouts is an essential ingredient of graphic design. The ability to condition layout generation on a partial subset of component attributes is critical to real-world applications that involve user interaction. Recently, diffusion models have demonstrated high-quality generative performances in various domains. However, it is unclear how to apply diffusion models to the natural representation of layouts which consists of a mix of discrete (class) and continuous (location, size) attributes. To address the conditioning layout generation problem, we introduce DLT, a joint discrete-continuous diffusion model. DLT is a transformer-based model which has a flexible conditioning mechanism that allows for conditioning on any given subset of all the layout component classes, locations, and sizes. Our method outperforms state-of-the-art generative models on various layout generation datasets with respect to different metrics and conditioning settings. Additionally, we validate the effectiveness of our proposed conditioning mechanism and the joint continuous-diffusion process. This joint process can be incorporated into a wide range of mixed discrete-continuous generative tasks.
☆ Zeroth-Order Optimization Meets Human Feedback: Provable Learning via Ranking Oracles
In this paper, we focus on a novel optimization problem in which the objective function is a black-box and can only be evaluated through a ranking oracle. This problem is common in real-world applications, particularly in cases where the function is assessed by human judges. Reinforcement Learning with Human Feedback (RLHF) is a prominent example of such an application, which is adopted by the recent works \cite{ouyang2022training,liu2023languages,chatgpt,bai2022training} to improve the quality of Large Language Models (LLMs) with human guidance. We propose ZO-RankSGD, a first-of-its-kind zeroth-order optimization algorithm, to solve this optimization problem with a theoretical guarantee. Specifically, our algorithm employs a new rank-based random estimator for the descent direction and is proven to converge to a stationary point. ZO-RankSGD can also be directly applied to the policy search problem in reinforcement learning when only a ranking oracle of the episode reward is available. This makes ZO-RankSGD a promising alternative to existing RLHF methods, as it optimizes in an online fashion and thus can work without any pre-collected data. Furthermore, we demonstrate the effectiveness of ZO-RankSGD in a novel application: improving the quality of images generated by a diffusion generative model with human ranking feedback. Throughout experiments, we found that ZO-RankSGD can significantly enhance the detail of generated images with only a few rounds of human feedback. Overall, our work advances the field of zeroth-order optimization by addressing the problem of optimizing functions with only ranking feedback, and offers an effective approach for aligning human and machine intentions in a wide range of domains. Our code is released here \url{https://github.com/TZW1998/Taming-Stable-Diffusion-with-Human-Ranking-Feedback}.
☆ Computing formation enthalpies through an explainable machine learning method: the case of Lanthanide Orthophosphates solid solutions
In the last decade, the use of Machine and Deep Learning (MDL) methods in Condensed Matter physics has seen a steep increase in the number of problems tackled and methods employed. A number of distinct MDL approaches have been employed in many different topics; from prediction of materials properties to computation of Density Functional Theory potentials and inter-atomic force fields. In many cases the result is a surrogate model which returns promising predictions but is opaque on the inner mechanisms of its success. On the other hand, the typical practitioner looks for answers that are explainable and provide a clear insight on the mechanisms governing a physical phenomena. In this work, we describe a proposal to use a sophisticated combination of traditional Machine Learning methods to obtain an explainable model that outputs an explicit functional formulation for the material property of interest. We demonstrate the effectiveness of our methodology in deriving a new highly accurate expression for the enthalpy of formation of solid solutions of lanthanides orthophosphates.
comment: 24 pages, 5 figures
☆ Graph Decision Transformer
Offline reinforcement learning (RL) is a challenging task, whose objective is to learn policies from static trajectory data without interacting with the environment. Recently, offline RL has been viewed as a sequence modeling problem, where an agent generates a sequence of subsequent actions based on a set of static transition experiences. However, existing approaches that use transformers to attend to all tokens naively can overlook the dependencies between different tokens and limit long-term dependency learning. In this paper, we propose the Graph Decision Transformer (GDT), a novel offline RL approach that models the input sequence into a causal graph to capture potential dependencies between fundamentally different concepts and facilitate temporal and causal relationship learning. GDT uses a graph transformer to process the graph inputs with relation-enhanced mechanisms, and an optional sequence transformer to handle fine-grained spatial information in visual tasks. Our experiments show that GDT matches or surpasses the performance of state-of-the-art offline RL methods on image-based Atari and OpenAI Gym.
comment: 14 pages
☆ High-Precision Machine-Learning Based Indoor Localization with Massive MIMO System
High-precision cellular-based localization is one of the key technologies for next-generation communication systems. In this paper, we investigate the potential of applying machine learning (ML) to a massive multiple-input multiple-output (MIMO) system to enhance localization accuracy. We analyze a new ML-based localization pipeline that has two parallel fully connected neural networks (FCNN). The first FCNN takes the instantaneous spatial covariance matrix to capture angular information, while the second FCNN takes the channel impulse responses to capture delay information. We fuse the estimated coordinates of these two FCNNs for further accuracy improvement. To test the localization algorithm, we performed an indoor measurement campaign with a massive MIMO testbed at 3.7GHz. In the measured scenario, the proposed pipeline can achieve centimeter-level accuracy by combining delay and angular information.
comment: 6 pages, 8 figures
☆ Multi-Dimensional and Multi-Scale Modeling for Speech Separation Optimized by Discriminative Learning ICASSP 2023
Transformer has shown advanced performance in speech separation, benefiting from its ability to capture global features. However, capturing local features and channel information of audio sequences in speech separation is equally important. In this paper, we present a novel approach named Intra-SE-Conformer and Inter-Transformer (ISCIT) for speech separation. Specifically, we design a new network SE-Conformer that can model audio sequences in multiple dimensions and scales, and apply it to the dual-path speech separation framework. Furthermore, we propose Multi-Block Feature Aggregation to improve the separation effect by selectively utilizing information from the intermediate blocks of the separation network. Meanwhile, we propose a speaker similarity discriminative loss to optimize the speech separation model to address the problem of poor performance when speakers have similar voices. Experimental results on the benchmark datasets WSJ0-2mix and WHAM! show that ISCIT can achieve state-of-the-art results.
comment: Accepted by ICASSP 2023
☆ A Multi-Stage Triple-Path Method for Speech Separation in Noisy and Reverberant Environments ICASSP 2023
In noisy and reverberant environments, the performance of deep learning-based speech separation methods drops dramatically because previous methods are not designed and optimized for such situations. To address this issue, we propose a multi-stage end-to-end learning method that decouples the difficult speech separation problem in noisy and reverberant environments into three sub-problems: speech denoising, separation, and de-reverberation. The probability and speed of searching for the optimal solution of the speech separation model are improved by reducing the solution space. Moreover, since the channel information of the audio sequence in the time domain is crucial for speech separation, we propose a triple-path structure capable of modeling the channel dimension of audio sequences. Experimental results show that the proposed multi-stage triple-path method can improve the performance of speech separation models at the cost of little model parameter increment.
comment: Accepted by ICASSP 2023
☆ Learning Bipedal Walking for Humanoids with Current Feedback IROS 2023
Recent advances in deep reinforcement learning (RL) based techniques combined with training in simulation have offered a new approach to developing control policies for legged robots. However, the application of such approaches to real hardware has largely been limited to quadrupedal robots with direct-drive actuators and light-weight bipedal robots with low gear-ratio transmission systems. Application to life-sized humanoid robots has been elusive due to the large sim-to-real gap arising from their large size, heavier limbs, and a high gear-ratio transmission systems. In this paper, we present an approach for effectively overcoming the sim-to-real gap issue for humanoid robots arising from inaccurate torque tracking at the actuator level. Our key idea is to utilize the current feedback from the motors on the real robot, after training the policy in a simulation environment artificially degraded with poor torque tracking. Our approach successfully trains an end-to-end policy in simulation that can be deployed on a real HRP-5P humanoid robot for bipedal locomotion on challenging terrain. We also perform robustness tests on the RL policy and compare its performance against a conventional model-based controller for walking on uneven terrain. YouTube video: https://youtu.be/IeUaSsBRbNY
comment: Submitted to the 2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2023). YouTube video: https://youtu.be/IeUaSsBRbNY
☆ Generative Modeling with Flow-Guided Density Ratio Learning
We present Flow-Guided Density Ratio Learning (FDRL), a simple and scalable approach to generative modeling which builds on the stale (time-independent) approximation of the gradient flow of entropy-regularized f-divergences introduced in DGflow. In DGflow, the intractable time-dependent density ratio is approximated by a stale estimator given by a GAN discriminator. This is sufficient in the case of sample refinement, where the source and target distributions of the flow are close to each other. However, this assumption is invalid for generation and a naive application of the stale estimator fails due to the large chasm between the two distributions. FDRL proposes to train a density ratio estimator such that it learns from progressively improving samples during the training process. We show that this simple method alleviates the density chasm problem, allowing FDRL to generate images of dimensions as high as $128\times128$, as well as outperform existing gradient flow baselines on quantitative benchmarks. We also show the flexibility of FDRL with two use cases. First, unconditional FDRL can be easily composed with external classifiers to perform class-conditional generation. Second, FDRL can be directly applied to unpaired image-to-image translation with no modifications needed to the framework. Code is publicly available at https://github.com/ajrheng/FDRL.
☆ Hybrid quantum-classical convolutional neural network for phytoplankton classification
The taxonomic composition and abundance of phytoplankton, having direct impact on marine ecosystem dynamic and global environment change, are listed as essential ocean variables. Phytoplankton classification is very crucial for Phytoplankton analysis, but it is very difficult because of the huge amount and tiny volume of Phytoplankton. Machine learning is the principle way of performing phytoplankton image classification automatically. When carrying out large-scale research on the marine phytoplankton, the volume of data increases overwhelmingly and more powerful computational resources are required for the success of machine learning algorithms. Recently, quantum machine learning has emerged as the potential solution for large-scale data processing by harnessing the exponentially computational power of quantum computer. Here, for the first time, we demonstrate the feasibility of quantum deep neural networks for phytoplankton classification. Hybrid quantum-classical convolutional and residual neural networks are developed based on the classical architectures. These models make a proper balance between the limited function of the current quantum devices and the large size of phytoplankton images, which make it possible to perform phytoplankton classification on the near-term quantum computers. Better performance is obtained by the quantum-enhanced models against the classical counterparts. In particular, quantum models converge much faster than classical ones. The present quantum models are versatile, and can be applied for various tasks of image classification in the field of marine science.
comment: 20 pages, 13 figures
☆ Classifying Text-Based Conspiracy Tweets related to COVID-19 using Contextualized Word Embeddings
The FakeNews task in MediaEval 2022 investigates the challenge of finding accurate and high-performance models for the classification of conspiracy tweets related to COVID-19. In this paper, we used BERT, ELMO, and their combination for feature extraction and RandomForest as classifier. The results show that ELMO performs slightly better than BERT, however their combination at feature level reduces the performance.
comment: Published in Multimedia Benchmark Workshop 2022, Bergen, Norway and Online, 12-13 January 2023: https://2022.multimediaeval.com/
☆ Variational Inference for Neyman-Scott Processes
Neyman-Scott processes (NSPs) have been applied across a range of fields to model points or temporal events with a hierarchy of clusters. Markov chain Monte Carlo (MCMC) is typically used for posterior sampling in the model. However, MCMC's mixing time can cause the resulting inference to be slow, and thereby slow down model learning and prediction. We develop the first variational inference (VI) algorithm for NSPs, and give two examples of suitable variational posterior point process distributions. Our method minimizes the inclusive Kullback-Leibler (KL) divergence for VI to obtain the variational parameters. We generate samples from the approximate posterior point processes much faster than MCMC, as we can directly estimate the approximate posterior point processes without any MCMC steps or gradient descent. We include synthetic and real-world data experiments that demonstrate our VI algorithm achieves better prediction performance than MCMC when computational time is limited.
☆ Stylometric Detection of AI-Generated Text in Twitter Timelines
Recent advancements in pre-trained language models have enabled convenient methods for generating human-like text at a large scale. Though these generation capabilities hold great potential for breakthrough applications, it can also be a tool for an adversary to generate misinformation. In particular, social media platforms like Twitter are highly susceptible to AI-generated misinformation. A potential threat scenario is when an adversary hijacks a credible user account and incorporates a natural language generator to generate misinformation. Such threats necessitate automated detectors for AI-generated tweets in a given user's Twitter timeline. However, tweets are inherently short, thus making it difficult for current state-of-the-art pre-trained language model-based detectors to accurately detect at what point the AI starts to generate tweets in a given Twitter timeline. In this paper, we present a novel algorithm using stylometric signals to aid detecting AI-generated tweets. We propose models corresponding to quantifying stylistic changes in human and AI tweets in two related tasks: Task 1 - discriminate between human and AI-generated tweets, and Task 2 - detect if and when an AI starts to generate tweets in a given Twitter timeline. Our extensive experiments demonstrate that the stylometric features are effective in augmenting the state-of-the-art AI-generated text detectors.
☆ MAST: Masked Augmentation Subspace Training for Generalizable Self-Supervised Priors ICLR 2023
Recent Self-Supervised Learning (SSL) methods are able to learn feature representations that are invariant to different data augmentations, which can then be transferred to downstream tasks of interest. However, different downstream tasks require different invariances for their best performance, so the optimal choice of augmentations for SSL depends on the target task. In this paper, we aim to learn self-supervised features that generalize well across a variety of downstream tasks (e.g., object classification, detection and instance segmentation) without knowing any task information beforehand. We do so by Masked Augmentation Subspace Training (or MAST) to encode in the single feature space the priors from different data augmentations in a factorized way. Specifically, we disentangle the feature space into separate subspaces, each induced by a learnable mask that selects relevant feature dimensions to model invariance to a specific augmentation. We show the success of MAST in jointly capturing generalizable priors from different augmentations, using both unique and shared features across the subspaces. We further show that MAST benefits from uncertainty modeling to reweight ambiguous samples from strong augmentations that may cause similarity mismatch in each subspace. Experiments demonstrate that MAST consistently improves generalization on various downstream tasks, while being task-agnostic and efficient during SSL. We also provide interesting insights about how different augmentations are related and how uncertainty reflects learning difficulty.
comment: ICLR 2023
☆ A Comparative Study of Deep Learning and Iterative Algorithms for Joint Channel Estimation and Signal Detection
Joint channel estimation and signal detection (JCESD) is crucial in wireless communication systems, but traditional algorithms perform poorly in low signal-to-noise ratio (SNR) scenarios. Deep learning (DL) methods have been investigated, but concerns regarding computational expense and lack of validation in low-SNR settings remain. Hence, the development of a robust and low-complexity model that can deliver excellent performance across a wide range of SNRs is highly desirable. In this paper, we aim to establish a benchmark where traditional algorithms and DL methods are validated on different channel models, Doppler, and SNR settings. In particular, we propose a new DL model where the backbone network is formed by unrolling the iterative algorithm, and the hyperparameters are estimated by hypernetworks. Additionally, we adapt a lightweight DenseNet to the task of JCESD for comparison. We evaluate different methods in three aspects: generalization in terms of bit error rate (BER), robustness, and complexity. Our results indicate that DL approaches outperform traditional algorithms in the challenging low-SNR setting, while the iterative algorithm performs better in highSNR settings. Furthermore, the iterative algorithm is more robust in the presence of carrier frequency offset, whereas DL methods excel when signals are corrupted by asymmetric Gaussian noise.
comment: 25 pages, this work has been submitted to the IEEE for possible publication. Code is available at https://github.com/j991222/MIMO_JCESD
☆ Training Machine Learning Models to Characterize Temporal Evolution of Disadvantaged Communities
Disadvantaged communities (DAC), as defined by the Justice40 initiative of the Department of Energy (DOE), USA, identifies census tracts across the USA to determine where benefits of climate and energy investments are or are not currently accruing. The DAC status not only helps in determining the eligibility for future Justice40-related investments but is also critical for exploring ways to achieve equitable distribution of resources. However, designing inclusive and equitable strategies not just requires a good understanding of current demographics, but also a deeper analysis of the transformations that happened in those demographics over the years. In this paper, machine learning (ML) models are trained on publicly available census data from recent years to classify the DAC status at the census tracts level and then the trained model is used to classify DAC status for historical years. A detailed analysis of the feature and model selection along with the evolution of disadvantaged communities between 2013 and 2018 is presented in this study.
☆ Face: Fast, Accurate and Context-Aware Audio Annotation and Classification
This paper presents a context-aware framework for feature selection and classification procedures to realize a fast and accurate audio event annotation and classification. The context-aware design starts with exploring feature extraction techniques to find an appropriate combination to select a set resulting in remarkable classification accuracy with minimal computational effort. The exploration for feature selection also embraces an investigation of audio Tempo representation, an advantageous feature extraction method missed by previous works in the environmental audio classification research scope. The proposed annotation method considers outlier, inlier, and hard-to-predict data samples to realize context-aware Active Learning, leading to the average accuracy of 90% when only 15% of data possess initial annotation. Our proposed algorithm for sound classification obtained average prediction accuracy of 98.05% on the UrbanSound8K dataset. The notebooks containing our source codes and implementation results are available at https://github.com/gitmehrdad/FACE.
comment: 9 pages, 4 figures
☆ ECG Classification System for Arrhythmia Detection Using Convolutional Neural Networks
Arrhythmia is just one of the many cardiovascular illnesses that have been extensively studied throughout the years. Using a multi-lead ECG data, this research describes a deep learning (DL) technique based on a convolutional neural network (CNN) algorithm to detect cardiovascular arrhythmia in patients. The suggested CNN model has six layers total, two convolution layers, two pooling layers, and two fully linked layers within a residual block, in addition to the input and output layers. In this study, the classification of the ECG signals into five groups, Left Bundle Branch Block (LBBB), Right Bundle Branch Block (RBBB), Atrial Premature Contraction (APC), Premature Ventricular Contraction (PVC), and Normal Beat is the main goal (N). Using the MIT-BIH arrhythmia dataset, we assessed the suggested technique. The findings show that our suggested strategy classified 15000 cases with an average accuracy of 98.2%.
☆ MPool: Motif-Based Graph Pooling
Graph Neural networks (GNNs) have recently become a powerful technique for many graph-related tasks including graph classification. Current GNN models apply different graph pooling methods that reduce the number of nodes and edges to learn the higher-order structure of the graph in a hierarchical way. All these methods primarily rely on the one-hop neighborhood. However, they do not consider the higher- order structure of the graph. In this work, we propose a multi-channel Motif-based Graph Pooling method named (MPool) captures the higher-order graph structure with motif and local and global graph structure with a combination of selection and clustering-based pooling operations. As the first channel, we develop node selection-based graph pooling by designing a node ranking model considering the motif adjacency of nodes. As the second channel, we develop cluster-based graph pooling by designing a spectral clustering model using motif adjacency. As the final layer, the result of each channel is aggregated into the final graph representation. We perform extensive experiments on eight benchmark datasets and show that our proposed method shows better accuracy than the baseline methods for graph classification tasks.
☆ Can Membership Inferencing be Refuted?
Membership inference (MI) attack is currently the most popular test for measuring privacy leakage in machine learning models. Given a machine learning model, a data point and some auxiliary information, the goal of an MI~attack is to determine whether the data point was used to train the model. In this work, we study the reliability of membership inference attacks in practice. Specifically, we show that a model owner can plausibly refute the result of a membership inference test on a data point $x$ by constructing a \textit{proof of repudiation} that proves that the model was trained \textit{without} $x$. We design efficient algorithms to construct proofs of repudiation for all data points of the training dataset. Our empirical evaluation demonstrates the practical feasibility of our algorithm by constructing proofs of repudiation for popular machine learning models on MNIST and CIFAR-10. Consequently, our results call for a re-evaluation of the implications of membership inference attacks in practice.
☆ AHPA: Adaptive Horizontal Pod Autoscaling Systems on Alibaba Cloud Container Service for Kubernetes
The existing resource allocation policy for application instances in Kubernetes cannot dynamically adjust according to the requirement of business, which would cause an enormous waste of resources during fluctuations. Moreover, the emergence of new cloud services puts higher resource management requirements. This paper discusses horizontal POD resources management in Alibaba Cloud Container Services with a newly deployed AI algorithm framework named AHPA -- the adaptive horizontal pod auto-scaling system. Based on a robust decomposition forecasting algorithm and performance training model, AHPA offers an optimal pod number adjustment plan that could reduce POD resources and maintain business stability. Since being deployed in April 2021, this system has expanded to multiple customer scenarios, including logistics, social networks, AI audio and video, e-commerce, etc. Compared with the previous algorithms, AHPA solves the elastic lag problem, increasing CPU usage by 10% and reducing resource cost by more than 20%. In addition, AHPA can automatically perform flexible planning according to the predicted business volume without manual intervention, significantly saving operation and maintenance costs.
☆ PreFallKD: Pre-Impact Fall Detection via CNN-ViT Knowledge Distillation
Fall accidents are critical issues in an aging and aged society. Recently, many researchers developed pre-impact fall detection systems using deep learning to support wearable-based fall protection systems for preventing severe injuries. However, most works only employed simple neural network models instead of complex models considering the usability in resource-constrained mobile devices and strict latency requirements. In this work, we propose a novel pre-impact fall detection via CNN-ViT knowledge distillation, namely PreFallKD, to strike a balance between detection performance and computational complexity. The proposed PreFallKD transfers the detection knowledge from the pre-trained teacher model (vision transformer) to the student model (lightweight convolutional neural networks). Additionally, we apply data augmentation techniques to tackle issues of data imbalance. We conduct the experiment on the KFall public dataset and compare PreFallKD with other state-of-the-art models. The experiment results show that PreFallKD could boost the student model during the testing phase and achieves reliable F1-score (92.66%) and lead time (551.3 ms).
☆ CoTEVer: Chain of Thought Prompting Annotation Toolkit for Explanation Verification EACL 2023
Chain-of-thought (CoT) prompting enables large language models (LLMs) to solve complex reasoning tasks by generating an explanation before the final prediction. Despite it's promising ability, a critical downside of CoT prompting is that the performance is greatly affected by the factuality of the generated explanation. To improve the correctness of the explanations, fine-tuning language models with explanation data is needed. However, there exists only a few datasets that can be used for such approaches, and no data collection tool for building them. Thus, we introduce CoTEVer, a tool-kit for annotating the factual correctness of generated explanations and collecting revision data of wrong explanations. Furthermore, we suggest several use cases where the data collected with CoTEVer can be utilized for enhancing the faithfulness of explanations. Our toolkit is publicly available at https://github.com/SeungoneKim/CoTEVer.
comment: Accepted at EACL 2023 Demo
☆ TinyAD: Memory-efficient anomaly detection for time series data in Industrial IoT
Monitoring and detecting abnormal events in cyber-physical systems is crucial to industrial production. With the prevalent deployment of the Industrial Internet of Things (IIoT), an enormous amount of time series data is collected to facilitate machine learning models for anomaly detection, and it is of the utmost importance to directly deploy the trained models on the IIoT devices. However, it is most challenging to deploy complex deep learning models such as Convolutional Neural Networks (CNNs) on these memory-constrained IIoT devices embedded with microcontrollers (MCUs). To alleviate the memory constraints of MCUs, we propose a novel framework named Tiny Anomaly Detection (TinyAD) to efficiently facilitate onboard inference of CNNs for real-time anomaly detection. First, we conduct a comprehensive analysis of depthwise separable CNNs and regular CNNs for anomaly detection and find that the depthwise separable convolution operation can reduce the model size by 50-90% compared with the traditional CNNs. Then, to reduce the peak memory consumption of CNNs, we explore two complementary strategies, in-place, and patch-by-patch memory rescheduling, and integrate them into a unified framework. The in-place method decreases the peak memory of the depthwise convolution by sparing a temporary buffer to transfer the activation results, while the patch-by-patch method further reduces the peak memory of layer-wise execution by slicing the input data into corresponding receptive fields and executing in order. Furthermore, by adjusting the dimension of convolution filters, these strategies apply to both univariate time series and multidomain time series features. Extensive experiments on real-world industrial datasets show that our framework can reduce peak memory consumption by 2-5x with negligible computation overhead.
comment: Accepted by IEEE Transactions on Industrial Informatics
☆ Adaptive Knowledge Distillation between Text and Speech Pre-trained Models
Learning on a massive amount of speech corpus leads to the recent success of many self-supervised speech models. With knowledge distillation, these models may also benefit from the knowledge encoded by language models that are pre-trained on rich sources of texts. The distillation process, however, is challenging due to the modal disparity between textual and speech embedding spaces. This paper studies metric-based distillation to align the embedding space of text and speech with only a small amount of data without modifying the model structure. Since the semantic and granularity gap between text and speech has been omitted in literature, which impairs the distillation, we propose the Prior-informed Adaptive knowledge Distillation (PAD) that adaptively leverages text/speech units of variable granularity and prior distributions to achieve better global and local alignments between text and speech pre-trained models. We evaluate on three spoken language understanding benchmarks to show that PAD is more effective in transferring linguistic knowledge than other metric-based distillation approaches.
☆ ADELT: Transpilation Between Deep Learning Frameworks
We propose Adversarial DEep Learning Transpiler (ADELT) for source-to-source transpilation between deep learning frameworks. Unlike prior approaches, we decouple the transpilation of code skeletons and the mapping of API keywords (an API function name or a parameter name). ADELT transpile code skeletons using few-shot prompting on big language models. Based on contextual embeddings extracted by a BERT for code, we train aligned API embeddings in a domain-adversarial setup, upon which we generate a dictionary for keyword translation. The model is trained on our unlabeled DL corpus from web crawl data, without using any hand-crafted rules and parallel data. Our method outperforms state-of-the-art transpilers on multiple transpilation pairs including PyTorch-Keras and PyTorch-MXNet by 15.9pts and 12.0pts in exact match scores respectively.
comment: 23 pages
☆ Exploring the Limits of Indiscriminate Data Poisoning Attacks
Indiscriminate data poisoning attacks aim to decrease a model's test accuracy by injecting a small amount of corrupted training data. Despite significant interest, existing attacks remain relatively ineffective against modern machine learning (ML) architectures. In this work, we introduce the notion of model poisonability as a technical tool to explore the intrinsic limits of data poisoning attacks. We derive an easily computable threshold to establish and quantify a surprising phase transition phenomenon among popular ML models: data poisoning attacks become effective only when the poisoning ratio exceeds our threshold. Building on existing parameter corruption attacks and refining the Gradient Canceling attack, we perform extensive experiments to confirm our theoretical findings, test the predictability of our transition threshold, and significantly improve existing data poisoning baselines over a range of datasets and models. Our work highlights the critical role played by the poisoning ratio, and sheds new insights on existing empirical results, attacks and mitigation strategies in data poisoning.
☆ Approach to Learning Generalized Audio Representation Through Batch Embedding Covariance Regularization and Constant-Q Transforms
General-purpose embedding is highly desirable for few-shot even zero-shot learning in many application scenarios, including audio tasks. In order to understand representations better, we conducted a thorough error analysis and visualization of HEAR 2021 submission results. Inspired by the analysis, this work experiments with different front-end audio preprocessing methods, including Constant-Q Transform (CQT) and Short-time Fourier transform (STFT), and proposes a Batch Embedding Covariance Regularization (BECR) term to uncover a more holistic simulation of the frequency information received by the human auditory system. We tested the models on the suite of HEAR 2021 tasks, which encompass a broad category of tasks. Preliminary results show (1) the proposed BECR can incur a more dispersed embedding on the test set, (2) BECR improves the PaSST model without extra computation complexity, and (3) STFT preprocessing outperforms CQT in all tasks we tested. Github:https://github.com/ankitshah009/general_audio_embedding_hear_2021
comment: Technical report, 10 pages
☆ Research on Efficient Fuzzy Clustering Method Based on Local Fuzzy Granules
In recent years, the problem of fuzzy clustering has been widely concerned. The membership iteration of existing methods is mostly considered globally, which has considerable problems in noisy environments, and iterative calculations for clusters with a large number of different sample sizes are not accurate and efficient. In this paper, starting from the strategy of large-scale priority, the data is fuzzy iterated using granular-balls, and the membership degree of data only considers the two granular-balls where it is located, thus improving the efficiency of iteration. The formed fuzzy granular-balls set can use more processing methods in the face of different data scenarios, which enhances the practicability of fuzzy clustering calculations.
☆ A Review of and Roadmap for Data Science and Machine Learning for the Neuropsychiatric Phenotype of Autism
Autism Spectrum Disorder (autism) is a neurodevelopmental delay which affects at least 1 in 44 children. Like many neurological disorder phenotypes, the diagnostic features are observable, can be tracked over time, and can be managed or even eliminated through proper therapy and treatments. Yet, there are major bottlenecks in the diagnostic, therapeutic, and longitudinal tracking pipelines for autism and related delays, creating an opportunity for novel data science solutions to augment and transform existing workflows and provide access to services for more affected families. Several prior efforts conducted by a multitude of research labs have spawned great progress towards improved digital diagnostics and digital therapies for children with autism. We review the literature of digital health methods for autism behavior quantification using data science. We describe both case-control studies and classification systems for digital phenotyping. We then discuss digital diagnostics and therapeutics which integrate machine learning models of autism-related behaviors, including the factors which must be addressed for translational use. Finally, we describe ongoing challenges and potent opportunities for the field of autism data science. Given the heterogeneous nature of autism and the complexities of the relevant behaviors, this review contains insights which are relevant to neurological behavior analysis and digital psychiatry more broadly.
☆ Learning When to Treat Business Processes: Prescriptive Process Monitoring with Causal Inference and Reinforcement Learning
Increasing the success rate of a process, i.e. the percentage of cases that end in a positive outcome, is a recurrent process improvement goal. At runtime, there are often certain actions (a.k.a. treatments) that workers may execute to lift the probability that a case ends in a positive outcome. For example, in a loan origination process, a possible treatment is to issue multiple loan offers to increase the probability that the customer takes a loan. Each treatment has a cost. Thus, when defining policies for prescribing treatments to cases, managers need to consider the net gain of the treatments. Also, the effect of a treatment varies over time: treating a case earlier may be more effective than later in a case. This paper presents a prescriptive monitoring method that automates this decision-making task. The method combines causal inference and reinforcement learning to learn treatment policies that maximize the net gain. The method leverages a conformal prediction technique to speed up the convergence of the reinforcement learning mechanism by separating cases that are likely to end up in a positive or negative outcome, from uncertain cases. An evaluation on two real-life datasets shows that the proposed method outperforms a state-of-the-art baseline.
☆ ERUDITE: Human-in-the-Loop IoT for an Adaptive Personalized Learning System
Thanks to the rapid growth in wearable technologies and recent advancement in machine learning and signal processing, monitoring complex human contexts becomes feasible, paving the way to develop human-in-the-loop IoT systems that naturally evolve to adapt to the human and environment state autonomously. Nevertheless, a central challenge in designing many of these IoT systems arises from the requirement to infer the human mental state, such as intention, stress, cognition load, or learning ability. While different human contexts can be inferred from the fusion of different sensor modalities that can correlate to a particular mental state, the human brain provides a richer sensor modality that gives us more insights into the required human context. This paper proposes ERUDITE, a human-in-the-loop IoT system for the learning environment that exploits recent wearable neurotechnology to decode brain signals. Through insights from concept learning theory, ERUDITE can infer the human state of learning and understand when human learning increases or declines. By quantifying human learning as an input sensory signal, ERUDITE can provide adequate personalized feedback to humans in a learning environment to enhance their learning experience. ERUDITE is evaluated across $15$ participants and showed that by using the brain signals as a sensor modality to infer the human learning state and providing personalized adaptation to the learning environment, the participants' learning performance increased on average by $26\%$. Furthermore, we showed that ERUDITE can be deployed on an edge-based prototype to evaluate its practicality and scalability.
comment: It is under review in the IEEE IoT journal
☆ Polynomial Time and Private Learning of Unbounded Gaussian Mixture Models
We study the problem of privately estimating the parameters of $d$-dimensional Gaussian Mixture Models (GMMs) with $k$ components. For this, we develop a technique to reduce the problem to its non-private counterpart. This allows us to privatize existing non-private algorithms in a blackbox manner, while incurring only a small overhead in the sample complexity and running time. As the main application of our framework, we develop an $(\varepsilon, \delta)$-differentially private algorithm to learn GMMs using the non-private algorithm of Moitra and Valiant [MV10] as a blackbox. Consequently, this gives the first sample complexity upper bound and first polynomial time algorithm for privately learning GMMs without any boundedness assumptions on the parameters.
☆ Sufficient dimension reduction for feature matrices
We address the problem of sufficient dimension reduction for feature matrices, which arises often in sensor network localization, brain neuroimaging, and electroencephalography analysis. In general, feature matrices have both row- and column-wise interpretations and contain structural information that can be lost with naive vectorization approaches. To address this, we propose a method called principal support matrix machine (PSMM) for the matrix sufficient dimension reduction. The PSMM converts the sufficient dimension reduction problem into a series of classification problems by dividing the response variables into slices. It effectively utilizes the matrix structure by finding hyperplanes with rank-1 normal matrix that optimally separate the sliced responses. Additionally, we extend our approach to the higher-order tensor case. Our numerical analysis demonstrates that the PSMM outperforms existing methods and has strong interpretability in real data applications.
comment: 30 pages, 3 figures
☆ CUDA: Convolution-based Unlearnable Datasets CVPR 2023
Large-scale training of modern deep learning models heavily relies on publicly available data on the web. This potentially unauthorized usage of online data leads to concerns regarding data privacy. Recent works aim to make unlearnable data for deep learning models by adding small, specially designed noises to tackle this issue. However, these methods are vulnerable to adversarial training (AT) and/or are computationally heavy. In this work, we propose a novel, model-free, Convolution-based Unlearnable DAtaset (CUDA) generation technique. CUDA is generated using controlled class-wise convolutions with filters that are randomly generated via a private key. CUDA encourages the network to learn the relation between filters and labels rather than informative features for classifying the clean data. We develop some theoretical analysis demonstrating that CUDA can successfully poison Gaussian mixture data by reducing the clean data performance of the optimal Bayes classifier. We also empirically demonstrate the effectiveness of CUDA with various datasets (CIFAR-10, CIFAR-100, ImageNet-100, and Tiny-ImageNet), and architectures (ResNet-18, VGG-16, Wide ResNet-34-10, DenseNet-121, DeIT, EfficientNetV2-S, and MobileNetV2). Our experiments show that CUDA is robust to various data augmentations and training approaches such as smoothing, AT with different budgets, transfer learning, and fine-tuning. For instance, training a ResNet-18 on ImageNet-100 CUDA achieves only 8.96$\%$, 40.08$\%$, and 20.58$\%$ clean test accuracies with empirical risk minimization (ERM), $L_{\infty}$ AT, and $L_{2}$ AT, respectively. Here, ERM on the clean training data achieves a clean test accuracy of 80.66$\%$. CUDA exhibits unlearnability effect with ERM even when only a fraction of the training dataset is perturbed. Furthermore, we also show that CUDA is robust to adaptive defenses designed specifically to break it.
comment: CVPR 2023
☆ A Computer Vision Enabled damage detection model with improved YOLOv5 based on Transformer Prediction Head
Objective:Computer vision-based up-to-date accurate damage classification and localization are of decisive importance for infrastructure monitoring, safety, and the serviceability of civil infrastructure. Current state-of-the-art deep learning (DL)-based damage detection models, however, often lack superior feature extraction capability in complex and noisy environments, limiting the development of accurate and reliable object distinction. Method: To this end, we present DenseSPH-YOLOv5, a real-time DL-based high-performance damage detection model where DenseNet blocks have been integrated with the backbone to improve in preserving and reusing critical feature information. Additionally, convolutional block attention modules (CBAM) have been implemented to improve attention performance mechanisms for strong and discriminating deep spatial feature extraction that results in superior detection under various challenging environments. Moreover, additional feature fusion layers and a Swin-Transformer Prediction Head (SPH) have been added leveraging advanced self-attention mechanism for more efficient detection of multiscale object sizes and simultaneously reducing the computational complexity. Results: Evaluating the model performance in large-scale Road Damage Dataset (RDD-2018), at a detection rate of 62.4 FPS, DenseSPH-YOLOv5 obtains a mean average precision (mAP) value of 85.25 %, F1-score of 81.18 %, and precision (P) value of 89.51 % outperforming current state-of-the-art models. Significance: The present research provides an effective and efficient damage localization model addressing the shortcoming of existing DL-based damage detection models by providing highly accurate localized bounding box prediction. Current work constitutes a step towards an accurate and robust automated damage detection system in real-time in-field applications.
☆ Amplitude-Varying Perturbation for Balancing Privacy and Utility in Federated Learning
While preserving the privacy of federated learning (FL), differential privacy (DP) inevitably degrades the utility (i.e., accuracy) of FL due to model perturbations caused by DP noise added to model updates. Existing studies have considered exclusively noise with persistent root-mean-square amplitude and overlooked an opportunity of adjusting the amplitudes to alleviate the adverse effects of the noise. This paper presents a new DP perturbation mechanism with a time-varying noise amplitude to protect the privacy of FL and retain the capability of adjusting the learning performance. Specifically, we propose a geometric series form for the noise amplitude and reveal analytically the dependence of the series on the number of global aggregations and the $(\epsilon,\delta)$-DP requirement. We derive an online refinement of the series to prevent FL from premature convergence resulting from excessive perturbation noise. Another important aspect is an upper bound developed for the loss function of a multi-layer perceptron (MLP) trained by FL running the new DP mechanism. Accordingly, the optimal number of global aggregations is obtained, balancing the learning and privacy. Extensive experiments are conducted using MLP, supporting vector machine, and convolutional neural network models on four public datasets. The contribution of the new DP mechanism to the convergence and accuracy of privacy-preserving FL is corroborated, compared to the state-of-the-art Gaussian noise mechanism with a persistent noise amplitude.
☆ On the Sample Complexity of Vanilla Model-Based Offline Reinforcement Learning with Dependent Samples AAAI-23
Offline reinforcement learning (offline RL) considers problems where learning is performed using only previously collected samples and is helpful for the settings in which collecting new data is costly or risky. In model-based offline RL, the learner performs estimation (or optimization) using a model constructed according to the empirical transition frequencies. We analyze the sample complexity of vanilla model-based offline RL with dependent samples in the infinite-horizon discounted-reward setting. In our setting, the samples obey the dynamics of the Markov decision process and, consequently, may have interdependencies. Under no assumption of independent samples, we provide a high-probability, polynomial sample complexity bound for vanilla model-based off-policy evaluation that requires partial or uniform coverage. We extend this result to the off-policy optimization under uniform coverage. As a comparison to the model-based approach, we analyze the sample complexity of off-policy evaluation with vanilla importance sampling in the infinite-horizon setting. Finally, we provide an estimator that outperforms the sample-mean estimator for almost deterministic dynamics that are prevalent in reinforcement learning.
comment: Accepted to AAAI-23
☆ adaPARL: Adaptive Privacy-Aware Reinforcement Learning for Sequential-Decision Making Human-in-the-Loop Systems
Reinforcement learning (RL) presents numerous benefits compared to rule-based approaches in various applications. Privacy concerns have grown with the widespread use of RL trained with privacy-sensitive data in IoT devices, especially for human-in-the-loop systems. On the one hand, RL methods enhance the user experience by trying to adapt to the highly dynamic nature of humans. On the other hand, trained policies can leak the user's private information. Recent attention has been drawn to designing privacy-aware RL algorithms while maintaining an acceptable system utility. A central challenge in designing privacy-aware RL, especially for human-in-the-loop systems, is that humans have intrinsic variability and their preferences and behavior evolve. The effect of one privacy leak mitigation can be different for the same human or across different humans over time. Hence, we can not design one fixed model for privacy-aware RL that fits all. To that end, we propose adaPARL, an adaptive approach for privacy-aware RL, especially for human-in-the-loop IoT systems. adaPARL provides a personalized privacy-utility trade-off depending on human behavior and preference. We validate the proposed adaPARL on two IoT applications, namely (i) Human-in-the-Loop Smart Home and (ii) Human-in-the-Loop Virtual Reality (VR) Smart Classroom. Results obtained on these two applications validate the generality of adaPARL and its ability to provide a personalized privacy-utility trade-off. On average, for the first application, adaPARL improves the utility by $57\%$ over the baseline and by $43\%$ over randomization. adaPARL also reduces the privacy leak by $23\%$ on average. For the second application, adaPARL decreases the privacy leak to $44\%$ before the utility drops by $15\%$.
comment: This paper is accepted at CPS-IoT week (IoTDI'23)
Self-supervised speech representation learning for keyword-spotting with light-weight transformers
Self-supervised speech representation learning (S3RL) is revolutionizing the way we leverage the ever-growing availability of data. While S3RL related studies typically use large models, we employ light-weight networks to comply with tight memory of compute-constrained devices. We demonstrate the effectiveness of S3RL on a keyword-spotting (KS) problem by using transformers with 330k parameters and propose a mechanism to enhance utterance-wise distinction, which proves crucial for improving performance on classification tasks. On the Google speech commands v2 dataset, the proposed method applied to the Auto-Regressive Predictive Coding S3RL led to a 1.2% accuracy improvement compared to training from scratch. On an in-house KS dataset with four different keywords, it provided 6% to 23.7% relative false accept improvement at fixed false reject rate. We argue this demonstrates the applicability of S3RL approaches to light-weight models for KS and confirms S3RL is a powerful alternative to traditional supervised learning for resource-constrained applications.
☆ TRACT: Denoising Diffusion Models with Transitive Closure Time-Distillation
Denoising Diffusion models have demonstrated their proficiency for generative sampling. However, generating good samples often requires many iterations. Consequently, techniques such as binary time-distillation (BTD) have been proposed to reduce the number of network calls for a fixed architecture. In this paper, we introduce TRAnsitive Closure Time-distillation (TRACT), a new method that extends BTD. For single step diffusion,TRACT improves FID by up to 2.4x on the same architecture, and achieves new single-step Denoising Diffusion Implicit Models (DDIM) state-of-the-art FID (7.4 for ImageNet64, 3.8 for CIFAR10). Finally we tease apart the method through extended ablations. The PyTorch implementation will be released soon.
☆ How Do Transformers Learn Topic Structure: Towards a Mechanistic Understanding
While the successes of transformers across many domains are indisputable, accurate understanding of the learning mechanics is still largely lacking. Their capabilities have been probed on benchmarks which include a variety of structured and reasoning tasks -- but mathematical understanding is lagging substantially behind. Recent lines of work have begun studying representational aspects of this question: that is, the size/depth/complexity of attention-based networks to perform certain tasks. However, there is no guarantee the learning dynamics will converge to the constructions proposed. In our paper, we provide fine-grained mechanistic understanding of how transformers learn "semantic structure", understood as capturing co-occurrence structure of words. Precisely, we show, through a combination of experiments on synthetic data modeled by Latent Dirichlet Allocation (LDA), Wikipedia data, and mathematical analysis that the embedding layer and the self-attention layer encode the topical structure. In the former case, this manifests as higher average inner product of embeddings between same-topic words. In the latter, it manifests as higher average pairwise attention between same-topic words. The mathematical results involve several assumptions to make the analysis tractable, which we verify on data, and might be of independent interest as well.
☆ A topological classifier to characterize brain states: When shape matters more than variance
Despite the remarkable accuracies attained by machine learning classifiers to separate complex datasets in a supervised fashion, most of their operation falls short to provide an informed intuition about the structure of data, and, what is more important, about the phenomena being characterized by the given datasets. By contrast, topological data analysis (TDA) is devoted to study the shape of data clouds by means of persistence descriptors and provides a quantitative characterization of specific topological features of the dataset under scrutiny. In this article we introduce a novel TDA-based classifier that works on the principle of assessing quantifiable changes on topological metrics caused by the addition of new input to a subset of data. We used this classifier with a high-dimensional electro-encephalographic (EEG) dataset recorded from eleven participants during a decision-making experiment in which three motivational states were induced through a manipulation of social pressure. After processing a band-pass filtered version of EEG signals, we calculated silhouettes from persistence diagrams associated with each motivated state, and classified unlabeled signals according to their impact on each reference silhouette. Our results show that in addition to providing accuracies within the range of those of a nearest neighbour classifier, the TDA classifier provides formal intuition of the structure of the dataset as well as an estimate of its intrinsic dimension. Towards this end, we incorporated dimensionality reduction methods to our procedure and found that the accuracy of our TDA classifier is generally not sensitive to explained variance but rather to shape, contrary to what happens with most machine learning classifiers.
comment: 21 pages, 13 figures
♻ ☆ Min-Max Bilevel Multi-objective Optimization with Applications in Machine Learning ICLR 2023
We consider a generic min-max multi-objective bilevel optimization problem with applications in robust machine learning such as representation learning and hyperparameter optimization. We design MORBiT, a novel single-loop gradient descent-ascent bilevel optimization algorithm, to solve the generic problem and present a novel analysis showing that MORBiT converges to the first-order stationary point at a rate of $\widetilde{\mathcal{O}}(n^{1/2} K^{-2/5})$ for a class of weakly convex problems with $n$ objectives upon $K$ iterations of the algorithm. Our analysis utilizes novel results to handle the non-smooth min-max multi-objective setup and to obtain a sublinear dependence in the number of objectives $n$. Experimental results on robust representation learning and robust hyperparameter optimization showcase (i) the advantages of considering the min-max multi-objective setup, and (ii) convergence properties of the proposed MORBiT. Our code is at https://github.com/minimario/MORBiT.
comment: 43 pages, 3 figures, ICLR 2023 version
♻ ☆ Boundary Graph Neural Networks for 3D Simulations AAAI
The abundance of data has given machine learning considerable momentum in natural sciences and engineering, though modeling of physical processes is often difficult. A particularly tough problem is the efficient representation of geometric boundaries. Triangularized geometric boundaries are well understood and ubiquitous in engineering applications. However, it is notoriously difficult to integrate them into machine learning approaches due to their heterogeneity with respect to size and orientation. In this work, we introduce an effective theory to model particle-boundary interactions, which leads to our new Boundary Graph Neural Networks (BGNNs) that dynamically modify graph structures to obey boundary conditions. The new BGNNs are tested on complex 3D granular flow processes of hoppers, rotating drums and mixers, which are all standard components of modern industrial machinery but still have complicated geometry. BGNNs are evaluated in terms of computational efficiency as well as prediction accuracy of particle flows and mixing entropies. BGNNs are able to accurately reproduce 3D granular flows within simulation uncertainties over hundreds of thousands of simulation timesteps. Most notably, in our experiments, particles stay within the geometric objects without using handcrafted conditions or restrictions.
comment: accepted for presentation at the Thirty-Seventh AAAI Conference on Artificial Intelligence (AAAI-23)
♻ ☆ Convergence under Lipschitz smoothness of ease-controlled Random Reshuffling gradient Algorithms
We consider minimizing the average of a very large number of smooth and possibly non-convex functions. This optimization problem has deserved much attention in the past years due to the many applications in different fields, the most challenging being training Machine Learning models. Widely used approaches for solving this problem are mini-batch gradient methods which, at each iteration, update the decision vector moving along the gradient of a mini-batch of the component functions. We consider the Incremental Gradient (IG) and the Random reshuffling (RR) methods which proceed in cycles, picking batches in a fixed order or by reshuffling the order after each epoch. Convergence properties of these schemes have been proved under different assumptions, usually quite strong. We aim to define ease-controlled modifications of the IG/RR schemes, which require a light additional computational effort and can be proved to converge under very weak and standard assumptions. In particular, we define two algorithmic schemes, monotone or non-monotone, in which the IG/RR iteration is controlled by using a watchdog rule and a derivative-free line search that activates only sporadically to guarantee convergence. The two schemes also allow controlling the updating of the stepsize used in the main IG/RR iteration, avoiding the use of preset rules. We prove convergence under the lonely assumption of Lipschitz continuity of the gradients of the component functions and perform extensive computational analysis using Deep Neural Architectures and a benchmark of datasets. We compare our implementation with both full batch gradient methods and online standard implementation of IG/RR methods, proving that the computational effort is comparable with the corresponding online methods and that the control on the learning rate may allow faster decrease.
comment: Add references, correct typos
♻ ☆ Optimum-statistical Collaboration Towards General and Efficient Black-box Optimization
In this paper, we make the key delineation on the roles of resolution and statistical uncertainty in hierarchical bandits-based black-box optimization algorithms, guiding a more general analysis and a more efficient algorithm design. We introduce the \textit{optimum-statistical collaboration}, an algorithm framework of managing the interaction between optimization error flux and statistical error flux evolving in the optimization process. We provide a general analysis of this framework without specifying the forms of statistical error and uncertainty quantifier. Our framework and its analysis, due to their generality, can be applied to a large family of functions and partitions that satisfy different local smoothness assumptions and have different numbers of local optimums, which is much richer than the class of functions studied in prior works. Our framework also inspires us to propose a better measure of the statistical uncertainty and consequently a variance-adaptive algorithm \texttt{VHCT}. In theory, we prove the algorithm enjoys rate-optimal regret bounds under different local smoothness assumptions; in experiments, we show the algorithm outperforms prior efforts in different settings.
♻ ☆ Prior and Posterior Networks: A Survey on Evidential Deep Learning Methods For Uncertainty Estimation
Popular approaches for quantifying predictive uncertainty in deep neural networks often involve distributions over weights or multiple models, for instance via Markov Chain sampling, ensembling, or Monte Carlo dropout. These techniques usually incur overhead by having to train multiple model instances or do not produce very diverse predictions. This comprehensive and extensive survey aims to familiarize the reader with an alternative class of models based on the concept of Evidential Deep Learning: For unfamiliar data, they aim to admit "what they don't know", and fall back onto a prior belief. Furthermore, they allow uncertainty estimation in a single model and forward pass by parameterizing distributions over distributions. This survey recapitulates existing works, focusing on the implementation in a classification setting, before surveying the application of the same paradigm to regression. We also reflect on the strengths and weaknesses compared to other existing methods and provide the most fundamental derivations using a unified notation to aid future research.
♻ ☆ Learning particle swarming models from data with Gaussian processes
Interacting particle or agent systems that display a rich variety of swarming behaviours are ubiquitous in science and engineering. A fundamental and challenging goal is to understand the link between individual interaction rules and swarming. In this paper, we study the data-driven discovery of a second-order particle swarming model that describes the evolution of $N$ particles in $\mathbb{R}^d$ under radial interactions. We propose a learning approach that models the latent radial interaction function as Gaussian processes, which can simultaneously fulfill two inference goals: one is the nonparametric inference of {the} interaction function with pointwise uncertainty quantification, and the other one is the inference of unknown scalar parameters in the non-collective friction forces of the system. We formulate the learning problem as a statistical inverse problem and provide a detailed analysis of recoverability conditions, establishing that a coercivity condition is sufficient for recoverability. Given data collected from $M$ i.i.d trajectories with independent Gaussian observational noise, we provide a finite-sample analysis, showing that our posterior mean estimator converges in a Reproducing kernel Hilbert space norm, at an optimal rate in $M$ equal to the one in the classical 1-dimensional Kernel Ridge regression. As a byproduct, we show we can obtain a parametric learning rate in $M$ for the posterior marginal variance using $L^{\infty}$ norm, and the rate could also involve $N$ and $L$ (the number of observation time instances for each trajectory), depending on the condition number of the inverse problem. Numerical results on systems that exhibit different swarming behaviors demonstrate efficient learning of our approach from scarce noisy trajectory data.
comment: 44 pages; Appendix 5 pages
♻ ☆ ACon$^2$: Adaptive Conformal Consensus for Provable Blockchain Oracles
Blockchains with smart contracts are distributed ledger systems that achieve block-state consistency among distributed nodes by only allowing deterministic operations of smart contracts. However, the power of smart contracts is enabled by interacting with stochastic off-chain data, which in turn opens the possibility to undermine the block-state consistency. To address this issue, an oracle smart contract is used to provide a single consistent source of external data; but, simultaneously, this introduces a single point of failure, which is called the oracle problem. To address the oracle problem, we propose an adaptive conformal consensus (ACon$^2$) algorithm that derives a consensus set of data from multiple oracle contracts via the recent advance in online uncertainty quantification learning. Interesting, the consensus set provides a desired correctness guarantee under distribution shift and Byzantine adversaries. We demonstrate the efficacy of the proposed algorithm on two price datasets and an Ethereum case study. In particular, the Solidity implementation of the proposed algorithm shows the potential practicality of the proposed algorithm, implying that online machine learning algorithms are applicable to address security issues in blockchains.
comment: Accepted to USENIX Security 2023
♻ ☆ Knowledge-augmented Graph Machine Learning for Drug Discovery: A Survey from Precision to Interpretability
The integration of Artificial Intelligence (AI) into the field of drug discovery has been a growing area of interdisciplinary scientific research. However, conventional AI models are heavily limited in handling complex biomedical structures (such as 2D or 3D protein and molecule structures) and providing interpretations for outputs, which hinders their practical application. As of late, Graph Machine Learning (GML) has gained considerable attention for its exceptional ability to model graph-structured biomedical data and investigate their properties and functional relationships. Despite extensive efforts, GML methods still suffer from several deficiencies, such as the limited ability to handle supervision sparsity and provide interpretability in learning and inference processes, and their ineffectiveness in utilising relevant domain knowledge. In response, recent studies have proposed integrating external biomedical knowledge into the GML pipeline to realise more precise and interpretable drug discovery with limited training instances. However, a systematic definition for this burgeoning research direction is yet to be established. This survey presents a comprehensive overview of long-standing drug discovery principles, provides the foundational concepts and cutting-edge techniques for graph-structured data and knowledge databases, and formally summarises Knowledge-augmented Graph Machine Learning (KaGML) for drug discovery. we propose a thorough review of related KaGML works, collected following a carefully designed search methodology, and organise them into four categories following a novel-defined taxonomy. To facilitate research in this promptly emerging field, we also share collected practical resources that are valuable for intelligent drug discovery and provide an in-depth discussion of the potential avenues for future advancements.
♻ ☆ On Calibrating Semantic Segmentation Models: Analyses and An Algorithm CVPR2023
We study the problem of semantic segmentation calibration. For image classification, lots of existing solutions are proposed to alleviate model miscalibration of confidence. However, to date, confidence calibration research on semantic segmentation is still limited. We provide a systematic study on the calibration of semantic segmentation models and propose a simple yet effective approach. First, we find that model capacity, crop size, multi-scale testing, and prediction correctness have impact on calibration. Among them, prediction correctness, especially misprediction, is more important to miscalibration due to over-confidence. Next, we propose a simple, unifying, and effective approach, namely selective scaling, by separating correct/incorrect prediction for scaling and more focusing on misprediction logit smoothing. Then, we study popular existing calibration methods and compare them with selective scaling on semantic segmentation calibration. We conduct extensive experiments with a variety of benchmarks on both in-domain and domain-shift calibration, and show that selective scaling consistently outperforms other methods.
comment: Accepted to CVPR2023 (8 pages, 4 figures)
♻ ☆ A Free Lunch from the Noise: Provable and Practical Exploration for Representation Learning UAI 2022
Representation learning lies at the heart of the empirical success of deep learning for dealing with the curse of dimensionality. However, the power of representation learning has not been fully exploited yet in reinforcement learning (RL), due to i), the trade-off between expressiveness and tractability; and ii), the coupling between exploration and representation learning. In this paper, we first reveal the fact that under some noise assumption in the stochastic control model, we can obtain the linear spectral feature of its corresponding Markov transition operator in closed-form for free. Based on this observation, we propose Spectral Dynamics Embedding (SPEDE), which breaks the trade-off and completes optimistic exploration for representation learning by exploiting the structure of the noise. We provide rigorous theoretical analysis of SPEDE, and demonstrate the practical superior performance over the existing state-of-the-art empirical algorithms on several benchmarks.
comment: UAI 2022. The first two authors contribute equally
♻ ☆ Learning Reward Functions for Robotic Manipulation by Observing Humans
Observing a human demonstrator manipulate objects provides a rich, scalable and inexpensive source of data for learning robotic policies. However, transferring skills from human videos to a robotic manipulator poses several challenges, not least a difference in action and observation spaces. In this work, we use unlabeled videos of humans solving a wide range of manipulation tasks to learn a task-agnostic reward function for robotic manipulation policies. Thanks to the diversity of this training data, the learned reward function sufficiently generalizes to image observations from a previously unseen robot embodiment and environment to provide a meaningful prior for directed exploration in reinforcement learning. We propose two methods for scoring states relative to a goal image: through direct temporal regression, and through distances in an embedding space obtained with time-contrastive learning. By conditioning the function on a goal image, we are able to reuse one model across a variety of tasks. Unlike prior work on leveraging human videos to teach robots, our method, Human Offline Learned Distances (HOLD) requires neither a priori data from the robot environment, nor a set of task-specific human demonstrations, nor a predefined notion of correspondence across morphologies, yet it is able to accelerate training of several manipulation tasks on a simulated robot arm compared to using only a sparse reward obtained from task completion.
♻ ☆ Latent Variable Representation for Reinforcement Learning ICLR 2023
Deep latent variable models have achieved significant empirical successes in model-based reinforcement learning (RL) due to their expressiveness in modeling complex transition dynamics. On the other hand, it remains unclear theoretically and empirically how latent variable models may facilitate learning, planning, and exploration to improve the sample efficiency of RL. In this paper, we provide a representation view of the latent variable models for state-action value functions, which allows both tractable variational learning algorithm and effective implementation of the optimism/pessimism principle in the face of uncertainty for exploration. In particular, we propose a computationally efficient planning algorithm with UCB exploration by incorporating kernel embeddings of latent variable models. Theoretically, we establish the sample complexity of the proposed approach in the online and offline settings. Empirically, we demonstrate superior performance over current state-of-the-art algorithms across various benchmarks.
comment: ICLR 2023. The first two authors contribute equally. Project Website: https://rlrep.github.io/lvrep/
♻ ☆ Spectral Decomposition Representation for Reinforcement Learning ICLR 2023
Representation learning often plays a critical role in reinforcement learning by managing the curse of dimensionality. A representative class of algorithms exploits a spectral decomposition of the stochastic transition dynamics to construct representations that enjoy strong theoretical properties in an idealized setting. However, current spectral methods suffer from limited applicability because they are constructed for state-only aggregation and derived from a policy-dependent transition kernel, without considering the issue of exploration. To address these issues, we propose an alternative spectral method, Spectral Decomposition Representation (SPEDER), that extracts a state-action abstraction from the dynamics without inducing spurious dependence on the data collection policy, while also balancing the exploration-versus-exploitation trade-off during learning. A theoretical analysis establishes the sample efficiency of the proposed algorithm in both the online and offline settings. In addition, an experimental investigation demonstrates superior performance over current state-of-the-art algorithms across several benchmarks.
comment: ICLR 2023. The first two authors contribute equally
♻ ☆ Optimal Methods for Convex Risk Averse Distributed Optimization
This paper studies the communication complexity of convex risk-averse optimization over a network. The problem generalizes the well-studied risk-neutral finite-sum distributed optimization problem and its importance stems from the need to handle risk in an uncertain environment. For algorithms in the literature, there exists a gap in communication complexities for solving risk-averse and risk-neutral problems. We propose two distributed algorithms, namely the distributed risk averse optimization (DRAO) method and the distributed risk averse optimization with sliding (DRAO-S) method, to close the gap. Specifically, the DRAO method achieves the optimal communication complexity by assuming a certain saddle point subproblem can be easily solved in the server node. The DRAO-S method removes the strong assumption by introducing a novel saddle point sliding subroutine which only requires the projection over the ambiguity set $P$. We observe that the number of $P$-projections performed by DRAO-S is optimal. Moreover, we develop matching lower complexity bounds to show the communication complexities of both DRAO and DRAO-S to be improvable. Numerical experiments are conducted to demonstrate the encouraging empirical performance of the DRAO-S method.
♻ ☆ Hybrid RL: Using Both Offline and Online Data Can Make RL Efficient ICLR 2023
We consider a hybrid reinforcement learning setting (Hybrid RL), in which an agent has access to an offline dataset and the ability to collect experience via real-world online interaction. The framework mitigates the challenges that arise in both pure offline and online RL settings, allowing for the design of simple and highly effective algorithms, in both theory and practice. We demonstrate these advantages by adapting the classical Q learning/iteration algorithm to the hybrid setting, which we call Hybrid Q-Learning or Hy-Q. In our theoretical results, we prove that the algorithm is both computationally and statistically efficient whenever the offline dataset supports a high-quality policy and the environment has bounded bilinear rank. Notably, we require no assumptions on the coverage provided by the initial distribution, in contrast with guarantees for policy gradient/iteration methods. In our experimental results, we show that Hy-Q with neural network function approximation outperforms state-of-the-art online, offline, and hybrid RL baselines on challenging benchmarks, including Montezuma's Revenge.
comment: 42 pages, 6 figures. Published at ICLR 2023. Code available at https://github.com/yudasong/HyQ
♻ ☆ Iterative Patch Selection for High-Resolution Image Recognition ICLR 2023
High-resolution images are prevalent in various applications, such as autonomous driving and computer-aided diagnosis. However, training neural networks on such images is computationally challenging and easily leads to out-of-memory errors even on modern GPUs. We propose a simple method, Iterative Patch Selection (IPS), which decouples the memory usage from the input size and thus enables the processing of arbitrarily large images under tight hardware constraints. IPS achieves this by selecting only the most salient patches, which are then aggregated into a global representation for image recognition. For both patch selection and aggregation, a cross-attention based transformer is introduced, which exhibits a close connection to Multiple Instance Learning. Our method demonstrates strong performance and has wide applicability across different domains, training regimes and image sizes while using minimal accelerator memory. For example, we are able to finetune our model on whole-slide images consisting of up to 250k patches (>16 gigapixels) with only 5 GB of GPU VRAM at a batch size of 16.
comment: Published as a conference paper at ICLR 2023
♻ ☆ Rapid training of quantum recurrent neural networks
Time series prediction is essential for human activities in diverse areas. A common approach to this task is to harness Recurrent Neural Networks (RNNs). However, while their predictions are quite accurate, their learning process is complex and, thus, time and energy consuming. Here, we propose to extend the concept of RRNs by including continuous-variable quantum resources in it, and to use a quantum-enhanced RNN to overcome these obstacles. The design of the Continuous-Variable Quantum RNN (CV-QRNN) is rooted in the continuous-variable quantum computing paradigm. By performing extensive numerical simulations, we demonstrate that the quantum network is capable of learning-time dependence of several types of temporal data, and that it converges to the optimal weights in fewer epochs than a classical network. Furthermore, for a small number of trainable parameters, it can achieve lower losses than its classical counterpart. CV-QRNN can be implemented using commercially available quantum-photonic hardware.
♻ ☆ Safe Testing
We develop the theory of hypothesis testing based on the e-value, a notion of evidence that, unlike the p-value, allows for effortlessly combining results from several studies in the common scenario where the decision to perform a new study may depend on previous outcomes. Tests based on e-values are safe, i.e. they preserve Type-I error guarantees, under such optional continuation. We define growth-rate optimality (GRO) as an analogue of power in an optional continuation context, and we show how to construct GRO e-variables for general testing problems with composite null and alternative, emphasizing models with nuisance parameters. GRO e-values take the form of Bayes factors with special priors. We illustrate the theory using several classic examples including a one-sample safe t-test and the 2 x 2 contingency table. Sharing Fisherian, Neymanian and Jeffreys-Bayesian interpretations, e-values may provide a methodology acceptable to adherents of all three schools.
comment: Accepted as discussion paper to the Journal of the Royal Statistical Society series B
♻ ☆ On the Importance of Sign Labeling: The Hamburg Sign Language Notation System Case Study
Labeling is the cornerstone of supervised machine learning, which has been exploited in a plethora of various applications, with sign language recognition being one of them. However, such algorithms must be fed with a huge amount of consistently labeled data during the training process to elaborate a well-generalizing model. In addition, there is a great need for an automated solution that works with any nationally diversified sign language. Although there are language-agnostic transcription systems, such as the Hamburg Sign Language Notation System (HamNoSys) that describe the signer's initial position and body movement instead of the glosses' meanings, there are still issues with providing accurate and reliable labels for every real-world use case. In this context, the industry relies heavily on manual attribution and labeling of the available video data. In this work, we tackle this issue and thoroughly analyze the HamNoSys labels provided by various maintainers of open sign language corpora in five sign languages, in order to examine the challenges encountered in labeling video data. We also investigate the consistency and objectivity of HamNoSys-based labels for the purpose of training machine learning models. Our findings provide valuable insights into the limitations of the current labeling methods and pave the way for future research on developing more accurate and efficient solutions for sign language recognition.
comment: 20 pages, 13 figures
♻ ☆ Grounding Graph Network Simulators using Physical Sensor Observations ICLR
Physical simulations that accurately model reality are crucial for many engineering disciplines such as mechanical engineering and robotic motion planning. In recent years, learned Graph Network Simulators produced accurate mesh-based simulations while requiring only a fraction of the computational cost of traditional simulators. Yet, the resulting predictors are confined to learning from data generated by existing mesh-based simulators and thus cannot include real world sensory information such as point cloud data. As these predictors have to simulate complex physical systems from only an initial state, they exhibit a high error accumulation for long-term predictions. In this work, we integrate sensory information to ground Graph Network Simulators on real world observations. In particular, we predict the mesh state of deformable objects by utilizing point cloud data. The resulting model allows for accurate predictions over longer time horizons, even under uncertainties in the simulation, such as unknown material properties. Since point clouds are usually not available for every time step, especially in online settings, we employ an imputation-based model. The model can make use of such additional information only when provided, and resorts to a standard Graph Network Simulator, otherwise. We experimentally validate our approach on a suite of prediction tasks for mesh-based interactions between soft and rigid bodies. Our method results in utilization of additional point cloud information to accurately predict stable simulations where existing Graph Network Simulators fail.
comment: Accepted as a poster at the 11th International Conference on Learning Representations (ICLR), 2023
♻ ☆ LambdaKG: A Library for Pre-trained Language Model-Based Knowledge Graph Embeddings
Knowledge Graphs (KGs) often have two characteristics: heterogeneous graph structure and text-rich entity/relation information. Text-based KG embeddings can represent entities by encoding descriptions with pre-trained language models, but no open-sourced library is specifically designed for KGs with PLMs at present. In this paper, we present LambdaKG, a library for KGE that equips with many pre-trained language models (e.g., BERT, BART, T5, GPT-3), and supports various tasks (e.g., knowledge graph completion, question answering, recommendation, and knowledge probing). LambdaKG is publicly open-sourced at https://github.com/zjunlp/PromptKG/tree/main/lambdaKG, with a demo video at http://deepke.zjukg.cn/lambdakg.mp4 and long-term maintenance.
comment: Work in progress and the project website is https://zjunlp.github.io/project/promptkg/
♻ ☆ Positive unlabeled learning with tensor networks
Positive unlabeled learning is a binary classification problem with positive and unlabeled data. It is common in domains where negative labels are costly or impossible to obtain, e.g., medicine and personalized advertising. We apply the locally purified state tensor network to the positive unlabeled learning problem and test our model on the MNIST image and 15 categorical/mixed datasets. On the MNIST dataset, we obtain close to the state-of-the-art results even with very few labeled positive samples. We significantly improve the state-of-the-art on categorical datasets. Further, we show that the agreement fraction between outputs of different models on unlabeled samples is a good indicator of the model's performance. Finally, our method can generate new positive and negative instances, which we demonstrate on simple synthetic datasets.
comment: 12 pages, 5 figures, 4 tables
♻ ☆ Denoising Masked AutoEncoders Help Robust Classification ICLR 2023
In this paper, we propose a new self-supervised method, which is called Denoising Masked AutoEncoders (DMAE), for learning certified robust classifiers of images. In DMAE, we corrupt each image by adding Gaussian noises to each pixel value and randomly masking several patches. A Transformer-based encoder-decoder model is then trained to reconstruct the original image from the corrupted one. In this learning paradigm, the encoder will learn to capture relevant semantics for the downstream tasks, which is also robust to Gaussian additive noises. We show that the pre-trained encoder can naturally be used as the base classifier in Gaussian smoothed models, where we can analytically compute the certified radius for any data point. Although the proposed method is simple, it yields significant performance improvement in downstream classification tasks. We show that the DMAE ViT-Base model, which just uses 1/10 parameters of the model developed in recent work arXiv:2206.10550, achieves competitive or better certified accuracy in various settings. The DMAE ViT-Large model significantly surpasses all previous results, establishing a new state-of-the-art on ImageNet dataset. We further demonstrate that the pre-trained model has good transferability to the CIFAR-10 dataset, suggesting its wide adaptability. Models and code are available at https://github.com/quanlin-wu/dmae.
comment: ICLR 2023
♻ ☆ Bayesian Neural Networks for Reversible Steganography
Recent advances in deep learning have led to a paradigm shift in the field of reversible steganography. A fundamental pillar of reversible steganography is predictive modelling which can be realised via deep neural networks. However, non-trivial errors exist in inferences about some out-of-distribution and noisy data. In view of this issue, we propose to consider uncertainty in predictive models based upon a theoretical framework of Bayesian deep learning, thereby creating an adaptive steganographic system. Most modern deep-learning models are regarded as deterministic because they only offer predictions while failing to provide uncertainty measurement. Bayesian neural networks bring a probabilistic perspective to deep learning and can be regarded as self-aware intelligent machinery; that is, a machine that knows its own limitations. To quantify uncertainty, we apply Bayesian statistics to model the predictive distribution and approximate it through Monte Carlo sampling with stochastic forward passes. We further show that predictive uncertainty can be disentangled into aleatoric and epistemic uncertainties and these quantities can be learnt unsupervised. Experimental results demonstrate an improvement delivered by Bayesian uncertainty analysis upon steganographic rate-distortion performance.
♻ ☆ UAV Path Planning Employing MPC- Reinforcement Learning Method Considering Collision Avoidance
In this paper, we tackle the problem of Unmanned Aerial (UA V) path planning in complex and uncertain environments by designing a Model Predictive Control (MPC), based on a Long-Short-Term Memory (LSTM) network integrated into the Deep Deterministic Policy Gradient algorithm. In the proposed solution, LSTM-MPC operates as a deterministic policy within the DDPG network, and it leverages a predicting pool to store predicted future states and actions for improved robustness and efficiency. The use of the predicting pool also enables the initialization of the critic network, leading to improved convergence speed and reduced failure rate compared to traditional reinforcement learning and deep reinforcement learning methods. The effectiveness of the proposed solution is evaluated by numerical simulations.
♻ ☆ Flow Annealed Importance Sampling Bootstrap
Normalizing flows are tractable density models that can approximate complicated target distributions, e.g. Boltzmann distributions of physical systems. However, current methods for training flows either suffer from mode-seeking behavior, use samples from the target generated beforehand by expensive MCMC methods, or use stochastic losses that have high variance. To avoid these problems, we augment flows with annealed importance sampling (AIS) and minimize the mass-covering $\alpha$-divergence with $\alpha=2$, which minimizes importance weight variance. Our method, Flow AIS Bootstrap (FAB), uses AIS to generate samples in regions where the flow is a poor approximation of the target, facilitating the discovery of new modes. We apply FAB to multimodal targets and show that we can approximate them very accurately where previous methods fail. To the best of our knowledge, we are the first to learn the Boltzmann distribution of the alanine dipeptide molecule using only the unnormalized target density, without access to samples generated via Molecular Dynamics (MD) simulations: FAB produces better results than training via maximum likelihood on MD samples while using 100 times fewer target evaluations. After reweighting the samples, we obtain unbiased histograms of dihedral angles that are almost identical to the ground truth.
♻ ☆ Bounding Information Leakage in Machine Learning
Recently, it has been shown that Machine Learning models can leak sensitive information about their training data. This information leakage is exposed through membership and attribute inference attacks. Although many attack strategies have been proposed, little effort has been made to formalize these problems. We present a novel formalism, generalizing membership and attribute inference attack setups previously studied in the literature and connecting them to memorization and generalization. First, we derive a universal bound on the success rate of inference attacks and connect it to the generalization gap of the target model. Second, we study the question of how much sensitive information is stored by the algorithm about its training set and we derive bounds on the mutual information between the sensitive attributes and model parameters. Experimentally, we illustrate the potential of our approach by applying it to both synthetic data and classification tasks on natural images. Finally, we apply our formalism to different attribute inference strategies, with which an adversary is able to recover the identity of writers in the PenDigits dataset.
comment: Published in [Elsevier Neurocomputing](https://doi.org/10.1016/j.neucom.2023.02.058)
♻ ☆ Survey of Machine Learning Based Intrusion Detection Methods for Internet of Medical Things
The Internet of Medical Things (IoMT) has revolutionized the healthcare industry by enabling physiological data collection using sensors, which are transmitted to remote servers for continuous analysis by physicians and healthcare professionals. This technology offers numerous benefits, including early disease detection and automatic medication for patients with chronic illnesses. However, IoMT technology also presents significant security risks, such as violating patient privacy or exposing sensitive data to interception attacks due to wireless communication, which could be fatal for the patient. Additionally, traditional security measures, such as cryptography, are challenging to implement in medical equipment due to the heterogeneous communication and their limited computation, storage, and energy capacity. These protection methods are also ineffective against new and zero-day attacks. It is essential to adopt robust security measures to ensure data integrity, confidentiality, and availability during data collection, transmission, storage, and processing. In this context, using Intrusion Detection Systems (IDS) based on Machine Learning (ML) can bring a complementary security solution adapted to the unique characteristics of IoMT systems. Therefore, this paper investigates how IDS based on ML can address security and privacy issues in IoMT systems. First, the generic three-layer architecture of IoMT is provided, and the security requirements of IoMT systems are outlined. Then, the various threats that can affect IoMT security are identified, and the advantages, disadvantages, methods, and datasets used in each solution based on ML at the three layers that make up IoMT are presented. Finally, the paper discusses the challenges and limitations of applying IDS based on ML at each layer of IoMT, which can serve as a future research direction.
comment: 40 pages, 3 figures, and 6 tables
♻ ☆ MHCCL: Masked Hierarchical Cluster-wise Contrastive Learning for Multivariate Time Series AAAI 2023
Learning semantic-rich representations from raw unlabeled time series data is critical for downstream tasks such as classification and forecasting. Contrastive learning has recently shown its promising representation learning capability in the absence of expert annotations. However, existing contrastive approaches generally treat each instance independently, which leads to false negative pairs that share the same semantics. To tackle this problem, we propose MHCCL, a Masked Hierarchical Cluster-wise Contrastive Learning model, which exploits semantic information obtained from the hierarchical structure consisting of multiple latent partitions for multivariate time series. Motivated by the observation that fine-grained clustering preserves higher purity while coarse-grained one reflects higher-level semantics, we propose a novel downward masking strategy to filter out fake negatives and supplement positives by incorporating the multi-granularity information from the clustering hierarchy. In addition, a novel upward masking strategy is designed in MHCCL to remove outliers of clusters at each partition to refine prototypes, which helps speed up the hierarchical clustering process and improves the clustering quality. We conduct experimental evaluations on seven widely-used multivariate time series datasets. The results demonstrate the superiority of MHCCL over the state-of-the-art approaches for unsupervised time series representation learning.
comment: accepted by AAAI 2023
♻ ☆ Perceive and predict: self-supervised speech representation based loss functions for speech enhancement ICASSP 2023
Recent work in the domain of speech enhancement has explored the use of self-supervised speech representations to aid in the training of neural speech enhancement models. However, much of this work focuses on using the deepest or final outputs of self supervised speech representation models, rather than the earlier feature encodings. The use of self supervised representations in such a way is often not fully motivated. In this work it is shown that the distance between the feature encodings of clean and noisy speech correlate strongly with psychoacoustically motivated measures of speech quality and intelligibility, as well as with human Mean Opinion Score (MOS) ratings. Experiments using this distance as a loss function are performed and improved performance over the use of STFT spectrogram distance based loss as well as other common loss functions from speech enhancement literature is demonstrated using objective measures such as perceptual evaluation of speech quality (PESQ) and short-time objective intelligibility (STOI).
comment: 4 pages, accepted at ICASSP 2023
♻ ☆ Model-Based Uncertainty in Value Functions AISTATS 2023
We consider the problem of quantifying uncertainty over expected cumulative rewards in model-based reinforcement learning. In particular, we focus on characterizing the variance over values induced by a distribution over MDPs. Previous work upper bounds the posterior variance over values by solving a so-called uncertainty Bellman equation, but the over-approximation may result in inefficient exploration. We propose a new uncertainty Bellman equation whose solution converges to the true posterior variance over values and explicitly characterizes the gap in previous work. Moreover, our uncertainty quantification technique is easily integrated into common exploration strategies and scales naturally beyond the tabular setting by using standard deep reinforcement learning architectures. Experiments in difficult exploration tasks, both in tabular and continuous control settings, show that our sharper uncertainty estimates improve sample-efficiency.
comment: AISTATS 2023
♻ ☆ Relative representations enable zero-shot latent space communication ICLR 2023
Neural networks embed the geometric structure of a data manifold lying in a high-dimensional space into latent representations. Ideally, the distribution of the data points in the latent space should depend only on the task, the data, the loss, and other architecture-specific constraints. However, factors such as the random weights initialization, training hyperparameters, or other sources of randomness in the training phase may induce incoherent latent spaces that hinder any form of reuse. Nevertheless, we empirically observe that, under the same data and modeling choices, the angles between the encodings within distinct latent spaces do not change. In this work, we propose the latent similarity between each sample and a fixed set of anchors as an alternative data representation, demonstrating that it can enforce the desired invariances without any additional training. We show how neural architectures can leverage these relative representations to guarantee, in practice, invariance to latent isometries and rescalings, effectively enabling latent space communication: from zero-shot model stitching to latent space comparison between diverse settings. We extensively validate the generalization capability of our approach on different datasets, spanning various modalities (images, text, graphs), tasks (e.g., classification, reconstruction) and architectures (e.g., CNNs, GCNs, transformers).
comment: ICLR 2023 notable top 5%, 26 pages, 11 figures, 18 tables
♻ ☆ Private, fair and accurate: Training large-scale, privacy-preserving AI models in medical imaging
Artificial intelligence (AI) models are increasingly used in the medical domain. However, as medical data is highly sensitive, special precautions to ensure its protection are required. The gold standard for privacy preservation is the introduction of differential privacy (DP) to model training. Prior work indicates that DP has negative implications on model accuracy and fairness, which are unacceptable in medicine and represent a main barrier to the widespread use of privacy-preserving techniques. In this work, we evaluated the effect of privacy-preserving training of AI models for chest radiograph diagnosis regarding accuracy and fairness compared to non-private training. For this, we used a large dataset (N=193,311) of high quality clinical chest radiographs, which were retrospectively collected and manually labeled by experienced radiologists. We then compared non-private deep convolutional neural networks (CNNs) and privacy-preserving (DP) models with respect to privacy-utility trade-offs measured as area under the receiver-operator-characteristic curve (AUROC), and privacy-fairness trade-offs, measured as Pearson's r or Statistical Parity Difference. We found that the non-private CNNs achieved an average AUROC score of 0.90 +- 0.04 over all labels, whereas the DP CNNs with a privacy budget of epsilon=7.89 resulted in an AUROC of 0.87 +- 0.04, i.e., a mere 2.6% performance decrease compared to non-private training. Furthermore, we found the privacy-preserving training not to amplify discrimination against age, sex or co-morbidity. Our study shows that -- under the challenging realistic circumstances of a real-life clinical dataset -- the privacy-preserving training of diagnostic deep learning models is possible with excellent diagnostic accuracy and fairness.
comment: 3 tables, 5 figures, 11 supplementary materials
♻ ☆ Online Low Rank Matrix Completion ICLR 2023
We study the problem of {\em online} low-rank matrix completion with $\mathsf{M}$ users, $\mathsf{N}$ items and $\mathsf{T}$ rounds. In each round, the algorithm recommends one item per user, for which it gets a (noisy) reward sampled from a low-rank user-item preference matrix. The goal is to design a method with sub-linear regret (in $\mathsf{T}$) and nearly optimal dependence on $\mathsf{M}$ and $\mathsf{N}$. The problem can be easily mapped to the standard multi-armed bandit problem where each item is an {\em independent} arm, but that leads to poor regret as the correlation between arms and users is not exploited. On the other hand, exploiting the low-rank structure of reward matrix is challenging due to non-convexity of the low-rank manifold. We first demonstrate that the low-rank structure can be exploited using a simple explore-then-commit (ETC) approach that ensures a regret of $O(\mathsf{polylog} (\mathsf{M}+\mathsf{N}) \mathsf{T}^{2/3})$. That is, roughly only $\mathsf{polylog} (\mathsf{M}+\mathsf{N})$ item recommendations are required per user to get a non-trivial solution. We then improve our result for the rank-$1$ setting which in itself is quite challenging and encapsulates some of the key issues. Here, we propose \textsc{OCTAL} (Online Collaborative filTering using iterAtive user cLustering) that guarantees nearly optimal regret of $O(\mathsf{polylog} (\mathsf{M}+\mathsf{N}) \mathsf{T}^{1/2})$. OCTAL is based on a novel technique of clustering users that allows iterative elimination of items and leads to a nearly optimal minimax rate.
comment: 37 pages, 7 figures (Accepted at ICLR 2023)
♻ ☆ Exploring Video Quality Assessment on User Generated Contents from Aesthetic and Technical Perspectives
The rapid increase in user-generated-content (UGC) videos calls for the development of effective video quality assessment (VQA) algorithms. However, the objective of the UGC-VQA problem is still ambiguous and can be viewed from two perspectives: the technical perspective, measuring the perception of distortions; and the aesthetic perspective, which relates to preference and recommendation on contents. To understand how these two perspectives affect overall subjective opinions in UGC-VQA, we conduct a large-scale subjective study to collect human quality opinions on overall quality of videos as well as perceptions from aesthetic and technical perspectives. The collected Disentangled Video Quality Database (DIVIDE-3k) confirms that human quality opinions on UGC videos are universally and inevitably affected by both aesthetic and technical perspectives. In light of this, we propose the Disentangled Objective Video Quality Evaluator (DOVER) to learn the quality of UGC videos based on the two perspectives. The DOVER proves state-of-the-art performance in UGC-VQA under very high efficiency. With perspective opinions in DIVIDE-3k, we further propose DOVER++, the first approach to provide reliable clear-cut quality evaluations from a single aesthetic or technical perspective. Code at https://github.com/VQAssessment/DOVER.
♻ ☆ Investigation of chemical structure recognition by encoder-decoder models in learning progress
Descriptor generation methods using latent representations of encoder$-$decoder (ED) models with SMILES as input are useful because of the continuity of descriptor and restorability to the structure. However, it is not clear how the structure is recognized in the learning progress of ED models. In this work, we created ED models of various learning progress and investigated the relationship between structural information and learning progress. We showed that compound substructures were learned early in ED models by monitoring the accuracy of downstream tasks and input$-$output substructure similarity using substructure$-$based descriptors, which suggests that existing evaluation methods based on the accuracy of downstream tasks may not be sensitive enough to evaluate the performance of ED models with SMILES as descriptor generation methods. On the other hand, we showed that structure restoration was time$-$consuming, and in particular, insufficient learning led to the estimation of a larger structure than the actual one. It can be inferred that determining the endpoint of the structure is a difficult task for the model. To our knowledge, this is the first study to link the learning progress of SMILES by ED model to chemical structures for a wide range of chemicals.
comment: 17 pages, 4 figures
♻ ☆ Learning Prototype-oriented Set Representations for Meta-Learning
Learning from set-structured data is a fundamental problem that has recently attracted increasing attention, where a series of summary networks are introduced to deal with the set input. In fact, many meta-learning problems can be treated as set-input tasks. Most existing summary networks aim to design different architectures for the input set in order to enforce permutation invariance. However, scant attention has been paid to the common cases where different sets in a meta-distribution are closely related and share certain statistical properties. Viewing each set as a distribution over a set of global prototypes, this paper provides a novel prototype-oriented optimal transport (POT) framework to improve existing summary networks. To learn the distribution over the global prototypes, we minimize its regularized optimal transport distance to the set empirical distribution over data points, providing a natural unsupervised way to improve the summary network. Since our plug-and-play framework can be applied to many meta-learning problems, we further instantiate it to the cases of few-shot classification and implicit meta generative modeling. Extensive experiments demonstrate that our framework significantly improves the existing summary networks on learning more powerful summary statistics from sets and can be successfully integrated into metric-based few-shot classification and generative modeling applications, providing a promising tool for addressing set-input and meta-learning problems.
♻ ☆ Using multimodal learning and deep generative models for corporate bankruptcy prediction
This research introduces for the first time, to the best of our knowledge, the concept of multimodal learning in bankruptcy prediction models. We use the Conditional Multimodal Discriminative (CMMD) model to learn multimodal representations that embed information from accounting, market, and textual modalities. The CMMD model needs a sample with all data modalities for model training. At test time, the CMMD model only needs access to accounting and market modalities to generate multimodal representations, which are further used to make bankruptcy predictions. This fact makes the use of bankruptcy prediction models using textual data realistic and possible, since accounting and market data are available for all companies unlike textual data. The empirical results in this research show that the classification performance of our proposed methodology is superior compared to that of a large number of traditional classifier models. We also show that our proposed methodology solves the limitation of previous bankruptcy models using textual data, as they can only make predictions for a small proportion of companies. Finally, based on multimodal representations, we introduce an index that is able to capture the uncertainty of the financial situation of companies during periods of financial distress.
♻ ☆ AutoTTS: End-to-End Text-to-Speech Synthesis through Differentiable Duration Modeling ICASSP 2023
Parallel text-to-speech (TTS) models have recently enabled fast and highly-natural speech synthesis. However, they typically require external alignment models, which are not necessarily optimized for the decoder as they are not jointly trained. In this paper, we propose a differentiable duration method for learning monotonic alignments between input and output sequences. Our method is based on a soft-duration mechanism that optimizes a stochastic process in expectation. Using this differentiable duration method, we introduce AutoTTS, a direct text-to-waveform speech synthesis model. AutoTTS enables high-fidelity speech synthesis through a combination of adversarial training and matching the total ground-truth duration. Experimental results show that our model obtains competitive results while enjoying a much simpler training pipeline. Audio samples are available online.
comment: ICASSP 2023
♻ ☆ Data Valuation Without Training of a Model ICLR 2023
Many recent works on understanding deep learning try to quantify how much individual data instances influence the optimization and generalization of a model. Such attempts reveal characteristics and importance of individual instances, which may provide useful information in diagnosing and improving deep learning. However, most of the existing works on data valuation require actual training of a model, which often demands high-computational cost. In this paper, we provide a training-free data valuation score, called complexity-gap score, which is a data-centric score to quantify the influence of individual instances in generalization of two-layer overparameterized neural networks. The proposed score can quantify irregularity of the instances and measure how much each data instance contributes in the total movement of the network parameters during training. We theoretically analyze and empirically demonstrate the effectiveness of the complexity-gap score in finding `irregular or mislabeled' data instances, and also provide applications of the score in analyzing datasets and diagnosing training dynamics. Our code is publicly available at https://github.com/JJchy/CG_score
comment: ICLR 2023
♻ ☆ Can discrete information extraction prompts generalize across language models? ICLR 2023
We study whether automatically-induced prompts that effectively extract information from a language model can also be used, out-of-the-box, to probe other language models for the same information. After confirming that discrete prompts induced with the AutoPrompt algorithm outperform manual and semi-manual prompts on the slot-filling task, we demonstrate a drop in performance for AutoPrompt prompts learned on a model and tested on another. We introduce a way to induce prompts by mixing language models at training time that results in prompts that generalize well across models. We conduct an extensive analysis of the induced prompts, finding that the more general prompts include a larger proportion of existing English words and have a less order-dependent and more uniform distribution of information across their component tokens. Our work provides preliminary evidence that it's possible to generate discrete prompts that can be induced once and used with a number of different models, and gives insights on the properties characterizing such prompts.
comment: Published as conference paper at ICLR 2023
♻ ☆ Learning-Assisted Algorithm Unrolling for Online Optimization with Budget Constraints AAAI'23
Online optimization with multiple budget constraints is challenging since the online decisions over a short time horizon are coupled together by strict inventory constraints. The existing manually-designed algorithms cannot achieve satisfactory average performance for this setting because they often need a large number of time steps for convergence and/or may violate the inventory constraints. In this paper, we propose a new machine learning (ML) assisted unrolling approach, called LAAU (Learning-Assisted Algorithm Unrolling), which unrolls the online decision pipeline and leverages an ML model for updating the Lagrangian multiplier online. For efficient training via backpropagation, we derive gradients of the decision pipeline over time. We also provide the average cost bounds for two cases when training data is available offline and collected online, respectively. Finally, we present numerical results to highlight that LAAU can outperform the existing baselines.
comment: Accepted by AAAI'23
♻ ☆ Learning Deep Semantics for Test Completion
Writing tests is a time-consuming yet essential task during software development. We propose to leverage recent advances in deep learning for text and code generation to assist developers in writing tests. We formalize the novel task of test completion to automatically complete the next statement in a test method based on the context of prior statements and the code under test. We develop TeCo -- a deep learning model using code semantics for test completion. The key insight underlying TeCo is that predicting the next statement in a test method requires reasoning about code execution, which is hard to do with only syntax-level data that existing code completion models use. TeCo extracts and uses six kinds of code semantics data, including the execution result of prior statements and the execution context of the test method. To provide a testbed for this new task, as well as to evaluate TeCo, we collect a corpus of 130,934 test methods from 1,270 open-source Java projects. Our results show that TeCo achieves an exact-match accuracy of 18, which is 29% higher than the best baseline using syntax-level data only. When measuring functional correctness of generated next statement, TeCo can generate runnable code in 29% of the cases compared to 18% obtained by the best baseline. Moreover, TeCo is significantly better than prior work on test oracle generation.
comment: Accepted as a conference paper in ICSE 2023
♻ ☆ GANStrument: Adversarial Instrument Sound Synthesis with Pitch-invariant Instance Conditioning ICASSP
We propose GANStrument, a generative adversarial model for instrument sound synthesis. Given a one-shot sound as input, it is able to generate pitched instrument sounds that reflect the timbre of the input within an interactive time. By exploiting instance conditioning, GANStrument achieves better fidelity and diversity of synthesized sounds and generalization ability to various inputs. In addition, we introduce an adversarial training scheme for a pitch-invariant feature extractor that significantly improves the pitch accuracy and timbre consistency. Experimental results show that GANStrument outperforms strong baselines that do not use instance conditioning in terms of generation quality and input editability. Qualitative examples are available online.
comment: 5 pages, 4 figures, Accepted to 2023 IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), Audio examples: https://ganstrument.github.io/ganstrument-demo/
♻ ☆ CLUTR: Curriculum Learning via Unsupervised Task Representation Learning
Reinforcement Learning (RL) algorithms are often known for sample inefficiency and difficult generalization. Recently, Unsupervised Environment Design (UED) emerged as a new paradigm for zero-shot generalization by simultaneously learning a task distribution and agent policies on the generated tasks. This is a non-stationary process where the task distribution evolves along with agent policies; creating an instability over time. While past works demonstrated the potential of such approaches, sampling effectively from the task space remains an open challenge, bottlenecking these approaches. To this end, we introduce CLUTR: a novel unsupervised curriculum learning algorithm that decouples task representation and curriculum learning into a two-stage optimization. It first trains a recurrent variational autoencoder on randomly generated tasks to learn a latent task manifold. Next, a teacher agent creates a curriculum by maximizing a minimax REGRET-based objective on a set of latent tasks sampled from this manifold. Using the fixed-pretrained task manifold, we show that CLUTR successfully overcomes the non-stationarity problem and improves stability. Our experimental results show CLUTR outperforms PAIRED, a principled and popular UED method, in the challenging CarRacing and navigation environments: achieving 10.6X and 45\% improvement in zero-shot generalization, respectively. CLUTR also performs comparably to the non-UED state-of-the-art for CarRacing, while requiring 500X fewer environment interactions.
comment: Preprint, Currently Under Review
♻ ☆ Discovery of Single Independent Latent Variable
Latent variable discovery is a central problem in data analysis with a broad range of applications in applied science. In this work, we consider data given as an invertible mixture of two statistically independent components and assume that one of the components is observed while the other is hidden. Our goal is to recover the hidden component. For this purpose, we propose an autoencoder equipped with a discriminator. Unlike the standard nonlinear ICA problem, which was shown to be non-identifiable, in the special case of ICA we consider here, we show that our approach can recover the component of interest up to entropy-preserving transformation. We demonstrate the performance of the proposed approach in several tasks, including image synthesis, voice cloning, and fetal ECG extraction.
comment: Published as a conference paper at Neurips 2022. In the current version the proof of the lemma is modified
♻ ☆ Deconstructed Generation-Based Zero-Shot Model AAAI 2023
Recent research on Generalized Zero-Shot Learning (GZSL) has focused primarily on generation-based methods. However, current literature has overlooked the fundamental principles of these methods and has made limited progress in a complex manner. In this paper, we aim to deconstruct the generator-classifier framework and provide guidance for its improvement and extension. We begin by breaking down the generator-learned unseen class distribution into class-level and instance-level distributions. Through our analysis of the role of these two types of distributions in solving the GZSL problem, we generalize the focus of the generation-based approach, emphasizing the importance of (i) attribute generalization in generator learning and (ii) independent classifier learning with partially biased data. We present a simple method based on this analysis that outperforms SotAs on four public GZSL datasets, demonstrating the validity of our deconstruction. Furthermore, our proposed method remains effective even without a generative model, representing a step towards simplifying the generator-classifier structure. Our code is available at \url{https://github.com/cdb342/DGZ}.
comment: AAAI 2023
♻ ☆ Rate-Optimal Contextual Online Matching Bandit
Two-sided online matching platforms have been employed in various markets. However, agents' preferences in present market are usually implicit and unknown and must be learned from data. With the growing availability of side information involved in the decision process, modern online matching methodology demands the capability to track preference dynamics for agents based on their contextual information. This motivates us to consider a novel Contextual Online Matching Bandit prOblem (COMBO), which allows dynamic preferences in matching decisions. Existing works focus on multi-armed bandit with static preference, but this is insufficient: the two-sided preference changes as along as one-side's contextual information updates, resulting in non-static matching. In this paper, we propose a Centralized Contextual - Explore Then Commit (CC-ETC) algorithm to adapt to the COMBO. CC-ETC solves online matching with dynamic preference. In theory, we show that CC-ETC achieves a sublinear regret upper bound O(log(T)) and is a rate-optimal algorithm by proving a matching lower bound. In the experiments, we demonstrate that CC-ETC is robust to variant preference schemes, dimensions of contexts, reward noise levels, and contexts variation levels.
♻ ☆ Bridging Distributional and Risk-sensitive Reinforcement Learning with Provable Regret Bounds
We study the regret guarantee for risk-sensitive reinforcement learning (RSRL) via distributional reinforcement learning (DRL) methods. In particular, we consider finite episodic Markov decision processes whose objective is the entropic risk measure (EntRM) of return. We identify a key property of the EntRM, the monotonicity-preserving property, which enables the risk-sensitive distributional dynamic programming framework. We then propose two novel DRL algorithms that implement optimism through two different schemes, including a model-free one and a model-based one. We prove that both of them attain $\tilde{\mathcal{O}}(\frac{\exp(|\beta| H)-1}{|\beta|H}H\sqrt{HS^2AT})$ regret upper bound, where $S$ is the number of states, $A$ the number of states, $H$ the time horizon and $T$ the number of total time steps. It matches RSVI2 proposed in \cite{fei2021exponential} with a much simpler regret analysis. To the best of our knowledge, this is the first regret analysis of DRL, which bridges DRL and RSRL in terms of sample complexity. Finally, we improve the existing lower bound by proving a tighter bound of $\Omega(\frac{\exp(\beta H/6)-1}{\beta H}H\sqrt{SAT})$ for $\beta>0$ case, which recovers the tight lower bound $\Omega(H\sqrt{SAT})$ in the risk-neutral setting.
♻ ☆ MLPInit: Embarrassingly Simple GNN Training Acceleration with MLP Initialization ICLR2023
Training graph neural networks (GNNs) on large graphs is complex and extremely time consuming. This is attributed to overheads caused by sparse matrix multiplication, which are sidestepped when training multi-layer perceptrons (MLPs) with only node features. MLPs, by ignoring graph context, are simple and faster for graph data, however they usually sacrifice prediction accuracy, limiting their applications for graph data. We observe that for most message passing-based GNNs, we can trivially derive an analog MLP (we call this a PeerMLP) with an equivalent weight space, by setting the trainable parameters with the same shapes, making us curious about \textbf{\emph{how do GNNs using weights from a fully trained PeerMLP perform?}} Surprisingly, we find that GNNs initialized with such weights significantly outperform their PeerMLPs, motivating us to use PeerMLP training as a precursor, initialization step to GNN training. To this end, we propose an embarrassingly simple, yet hugely effective initialization method for GNN training acceleration, called MLPInit. Our extensive experiments on multiple large-scale graph datasets with diverse GNN architectures validate that MLPInit can accelerate the training of GNNs (up to 33X speedup on OGB-Products) and often improve prediction performance (e.g., up to $7.97\%$ improvement for GraphSAGE across $7$ datasets for node classification, and up to $17.81\%$ improvement across $4$ datasets for link prediction on metric Hits@10). The code is available at \href{https://github.com/snap-research/MLPInit-for-GNNs}.
comment: Accepted by ICLR2023
♻ ☆ Parallel Deep Neural Networks Have Zero Duality Gap
Training deep neural networks is a challenging non-convex optimization problem. Recent work has proven that the strong duality holds (which means zero duality gap) for regularized finite-width two-layer ReLU networks and consequently provided an equivalent convex training problem. However, extending this result to deeper networks remains to be an open problem. In this paper, we prove that the duality gap for deeper linear networks with vector outputs is non-zero. In contrast, we show that the zero duality gap can be obtained by stacking standard deep networks in parallel, which we call a parallel architecture, and modifying the regularization. Therefore, we prove the strong duality and existence of equivalent convex problems that enable globally optimal training of deep networks. As a by-product of our analysis, we demonstrate that the weight decay regularization on the network parameters explicitly encourages low-rank solutions via closed-form expressions. In addition, we show that strong duality holds for three-layer standard ReLU networks given rank-1 data matrices.
♻ ☆ Physics-Informed Machine Learning: A Survey on Problems, Methods and Applications
Recent advances of data-driven machine learning have revolutionized fields like computer vision, reinforcement learning, and many scientific and engineering domains. In many real-world and scientific problems, systems that generate data are governed by physical laws. Recent work shows that it provides potential benefits for machine learning models by incorporating the physical prior and collected data, which makes the intersection of machine learning and physics become a prevailing paradigm. By integrating the data and mathematical physics models seamlessly, it can guide the machine learning model towards solutions that are physically plausible, improving accuracy and efficiency even in uncertain and high-dimensional contexts. In this survey, we present this learning paradigm called Physics-Informed Machine Learning (PIML) which is to build a model that leverages empirical data and available physical prior knowledge to improve performance on a set of tasks that involve a physical mechanism. We systematically review the recent development of physics-informed machine learning from three perspectives of machine learning tasks, representation of physical prior, and methods for incorporating physical prior. We also propose several important open research problems based on the current trends in the field. We argue that encoding different forms of physical prior into model architectures, optimizers, inference algorithms, and significant domain-specific applications like inverse engineering design and robotic control is far from being fully explored in the field of physics-informed machine learning. We believe that the interdisciplinary research of physics-informed machine learning will significantly propel research progress, foster the creation of more effective machine learning models, and also offer invaluable assistance in addressing long-standing problems in related disciplines.
♻ ☆ Tight Certification of Adversarially Trained Neural Networks via Nonconvex Low-Rank Semidefinite Relaxations
Adversarial training is well-known to produce high-quality neural network models that are empirically robust against adversarial perturbations. Nevertheless, once a model has been adversarially trained, one often desires a certification that the model is truly robust against all future attacks. Unfortunately, when faced with adversarially trained models, all existing approaches have significant trouble making certifications that are strong enough to be practically useful. Linear programming (LP) techniques in particular face a "convex relaxation barrier" that prevent them from making high-quality certifications, even after refinement with mixed-integer linear programming (MILP) techniques, and even when using state-of-the-art computational facilities. In this paper, we propose a nonconvex certification technique, based on a low-rank restriction of a semidefinite programming (SDP) relaxation. The nonconvex relaxation makes strong certifications comparable to much more expensive SDP methods, while optimizing over dramatically fewer variables comparable to much weaker LP methods. Despite nonconvexity, we show how off-the-shelf local optimization algorithms can be used to achieve and to certify global optimality in polynomial time. Our experiments find that the nonconvex relaxation almost completely closes the gap towards exact certification of adversarially trained models.
♻ ☆ Gluformer: Transformer-Based Personalized Glucose Forecasting with Uncertainty Quantification ICASSP
Deep learning models achieve state-of-the art results in predicting blood glucose trajectories, with a wide range of architectures being proposed. However, the adaptation of such models in clinical practice is slow, largely due to the lack of uncertainty quantification of provided predictions. In this work, we propose to model the future glucose trajectory conditioned on the past as an infinite mixture of basis distributions (i.e., Gaussian, Laplace, etc.). This change allows us to learn the uncertainty and predict more accurately in the cases when the trajectory has a heterogeneous or multi-modal distribution. To estimate the parameters of the predictive distribution, we utilize the Transformer architecture. We empirically demonstrate the superiority of our method over existing state-of-the-art techniques both in terms of accuracy and uncertainty on the synthetic and benchmark glucose data sets.
comment: 5 pages, 2 figures, IEEE ICASSP
♻ ☆ VIP: Towards Universal Visual Reward and Representation via Value-Implicit Pre-Training ICLR 2023
Reward and representation learning are two long-standing challenges for learning an expanding set of robot manipulation skills from sensory observations. Given the inherent cost and scarcity of in-domain, task-specific robot data, learning from large, diverse, offline human videos has emerged as a promising path towards acquiring a generally useful visual representation for control; however, how these human videos can be used for general-purpose reward learning remains an open question. We introduce $\textbf{V}$alue-$\textbf{I}$mplicit $\textbf{P}$re-training (VIP), a self-supervised pre-trained visual representation capable of generating dense and smooth reward functions for unseen robotic tasks. VIP casts representation learning from human videos as an offline goal-conditioned reinforcement learning problem and derives a self-supervised dual goal-conditioned value-function objective that does not depend on actions, enabling pre-training on unlabeled human videos. Theoretically, VIP can be understood as a novel implicit time contrastive objective that generates a temporally smooth embedding, enabling the value function to be implicitly defined via the embedding distance, which can then be used to construct the reward for any goal-image specified downstream task. Trained on large-scale Ego4D human videos and without any fine-tuning on in-domain, task-specific data, VIP's frozen representation can provide dense visual reward for an extensive set of simulated and $\textbf{real-robot}$ tasks, enabling diverse reward-based visual control methods and significantly outperforming all prior pre-trained representations. Notably, VIP can enable simple, $\textbf{few-shot}$ offline RL on a suite of real-world robot tasks with as few as 20 trajectories.
comment: ICLR 2023, Notable-Top-25% (Spotlight). Project website: https://sites.google.com/view/vip-rl
♻ ☆ A Deep Reinforcement Learning Approach for Finding Non-Exploitable Strategies in Two-Player Atari Games
This paper proposes new, end-to-end deep reinforcement learning algorithms for learning two-player zero-sum Markov games. Different from prior efforts on training agents to beat a fixed set of opponents, our objective is to find the Nash equilibrium policies that are free from exploitation by even the adversarial opponents. We propose (a) Nash-DQN algorithm, which integrates the deep learning techniques from single DQN into the classic Nash Q-learning algorithm for solving tabular Markov games; (b) Nash-DQN-Exploiter algorithm, which additionally adopts an exploiter to guide the exploration of the main agent. We conduct experimental evaluation on tabular examples as well as various two-player Atari games. Our empirical results demonstrate that (i) the policies found by many existing methods including Neural Fictitious Self Play and Policy Space Response Oracle can be prone to exploitation by adversarial opponents; (ii) the output policies of our algorithms are robust to exploitation, and thus outperform existing methods.
♻ ☆ Towards Memory- and Time-Efficient Backpropagation for Training Spiking Neural Networks
Spiking Neural Networks (SNNs) are promising energy-efficient models for neuromorphic computing. For training the non-differentiable SNN models, the backpropagation through time (BPTT) with surrogate gradients (SG) method has achieved high performance. However, this method suffers from considerable memory cost and training time during training. In this paper, we propose the Spatial Learning Through Time (SLTT) method that can achieve high performance while greatly improving training efficiency compared with BPTT. First, we show that the backpropagation of SNNs through the temporal domain contributes just a little to the final calculated gradients. Thus, we propose to ignore the unimportant routes in the computational graph during backpropagation. The proposed method reduces the number of scalar multiplications and achieves a small memory occupation that is independent of the total time steps. Furthermore, we propose a variant of SLTT, called SLTT-K, that allows backpropagation only at K time steps, then the required number of scalar multiplications is further reduced and is independent of the total time steps. Experiments on both static and neuromorphic datasets demonstrate superior training efficiency and performance of our SLTT. In particular, our method achieves state-of-the-art accuracy on ImageNet, while the memory cost and training time are reduced by more than 70% and 50%, respectively, compared with BPTT.
♻ ☆ Bootstrapping Semi-supervised Medical Image Segmentation with Anatomical-aware Contrastive Distillation
Contrastive learning has shown great promise over annotation scarcity problems in the context of medical image segmentation. Existing approaches typically assume a balanced class distribution for both labeled and unlabeled medical images. However, medical image data in reality is commonly imbalanced (i.e., multi-class label imbalance), which naturally yields blurry contours and usually incorrectly labels rare objects. Moreover, it remains unclear whether all negative samples are equally negative. In this work, we present ACTION, an Anatomical-aware ConTrastive dIstillatiON framework, for semi-supervised medical image segmentation. Specifically, we first develop an iterative contrastive distillation algorithm by softly labeling the negatives rather than binary supervision between positive and negative pairs. We also capture more semantically similar features from the randomly chosen negative set compared to the positives to enforce the diversity of the sampled data. Second, we raise a more important question: Can we really handle imbalanced samples to yield better performance? Hence, the key innovation in ACTION is to learn global semantic relationship across the entire dataset and local anatomical features among the neighbouring pixels with minimal additional memory footprint. During the training, we introduce anatomical contrast by actively sampling a sparse set of hard negative pixels, which can generate smoother segmentation boundaries and more accurate predictions. Extensive experiments across two benchmark datasets and different unlabeled settings show that ACTION significantly outperforms the current state-of-the-art semi-supervised methods.
♻ ☆ Reparameterization through Spatial Gradient Scaling ICLR 2023
Reparameterization aims to improve the generalization of deep neural networks by transforming convolutional layers into equivalent multi-branched structures during training. However, there exists a gap in understanding how reparameterization may change and benefit the learning process of neural networks. In this paper, we present a novel spatial gradient scaling method to redistribute learning focus among weights in convolutional networks. We prove that spatial gradient scaling achieves the same learning dynamics as a branched reparameterization yet without introducing structural changes into the network. We further propose an analytical approach that dynamically learns scalings for each convolutional layer based on the spatial characteristics of its input feature map gauged by mutual information. Experiments on CIFAR-10, CIFAR-100, and ImageNet show that without searching for reparameterized structures, our proposed scaling method outperforms the state-of-the-art reparameterization strategies at a lower computational cost.
comment: Published at ICLR 2023. Code available at https://github.com/Ascend-Research/Reparameterization
♻ ☆ NSGA-PINN: A Multi-Objective Optimization Method for Physics-Informed Neural Network Training
This paper presents NSGA-PINN, a multi-objective optimization framework for effective training of Physics-Informed Neural Networks (PINNs). The proposed framework uses the Non-dominated Sorting Genetic Algorithm (NSGA-II) to enable traditional stochastic gradient optimization algorithms (e.g., ADAM) to escape local minima effectively. Additionally, the NSGA-II algorithm enables satisfying the initial and boundary conditions encoded into the loss function during physics-informed training precisely. We demonstrate the effectiveness of our framework by applying NSGA-PINN to several ordinary and partial differential equation problems. In particular, we show that the proposed framework can handle challenging inverse problems with noisy data.
comment: 13 pages, 35 figures
♻ ☆ Tier Balancing: Towards Dynamic Fairness over Underlying Causal Factors
The pursuit of long-term fairness involves the interplay between decision-making and the underlying data generating process. In this paper, through causal modeling with a directed acyclic graph (DAG) on the decision-distribution interplay, we investigate the possibility of achieving long-term fairness from a dynamic perspective. We propose Tier Balancing, a technically more challenging but more natural notion to achieve in the context of long-term, dynamic fairness analysis. Different from previous fairness notions that are defined purely on observed variables, our notion goes one step further, capturing behind-the-scenes situation changes on the unobserved latent causal factors that directly carry out the influence from the current decision to the future data distribution. Under the specified dynamics, we prove that in general one cannot achieve the long-term fairness goal only through one-step interventions. Furthermore, in the effort of approaching long-term fairness, we consider the mission of "getting closer to" the long-term fairness goal and present possibility and impossibility results accordingly.
♻ ☆ Accelerate the Warm-up Stage in the Lasso Computation via a Homotopic Approach
In optimization, it is known that when the objective functions are strictly convex and well-conditioned, gradient-based approaches can be extremely effective, e.g., achieving the exponential rate of convergence. On the other hand, the existing Lasso-type estimator in general cannot achieve the optimal rate due to the undesirable behavior of the absolute function at the origin. A homotopic method is to use a sequence of surrogate functions to approximate the $\ell_1$ penalty that is used in the Lasso-type of estimators. The surrogate functions will converge to the $\ell_1$ penalty in the Lasso estimator. At the same time, each surrogate function is strictly convex, which enables a provable faster numerical rate of convergence. In this paper, we demonstrate that by meticulously defining the surrogate functions, one can prove a faster numerical convergence rate than any existing methods in computing for the Lasso-type of estimators. Namely, the state-of-the-art algorithms can only guarantee $O(1/\epsilon)$ or $O(1/\sqrt{\epsilon})$ convergence rates, while we can prove an $O([\log(1/\epsilon)]^2)$ for the newly proposed algorithm. Our numerical simulations show that the new algorithm also performs better empirically.
comment: 19 pages, 3 figures, 3 tables
♻ ☆ Riemannian Metric Learning via Optimal Transport ICLR 2023
We introduce an optimal transport-based model for learning a metric tensor from cross-sectional samples of evolving probability measures on a common Riemannian manifold. We neurally parametrize the metric as a spatially-varying matrix field and efficiently optimize our model's objective using a simple alternating scheme. Using this learned metric, we can nonlinearly interpolate between probability measures and compute geodesics on the manifold. We show that metrics learned using our method improve the quality of trajectory inference on scRNA and bird migration data at the cost of little additional cross-sectional data.
comment: ICLR 2023
♻ ☆ Low Budget Active Learning via Wasserstein Distance: An Integer Programming Approach
Active learning is the process of training a model with limited labeled data by selecting a core subset of an unlabeled data pool to label. The large scale of data sets used in deep learning forces most sample selection strategies to employ efficient heuristics. This paper introduces an integer optimization problem for selecting a core set that minimizes the discrete Wasserstein distance from the unlabeled pool. We demonstrate that this problem can be tractably solved with a Generalized Benders Decomposition algorithm. Our strategy uses high-quality latent features that can be obtained by unsupervised learning on the unlabeled pool. Numerical results on several data sets show that our optimization approach is competitive with baselines and particularly outperforms them in the low budget regime where less than one percent of the data set is labeled.
♻ ☆ Human-in-the-Loop Mixup
Aligning model representations to humans has been found to improve robustness and generalization. However, such methods often focus on standard observational data. Synthetic data is proliferating and powering many advances in machine learning; yet, it is not always clear whether synthetic labels are perceptually aligned to humans -- rendering it likely model representations are not human aligned. We focus on the synthetic data used in mixup: a powerful regularizer shown to improve model robustness, generalization, and calibration. We design a comprehensive series of elicitation interfaces, which we release as HILL MixE Suite, and recruit 159 participants to provide perceptual judgments along with their uncertainties, over mixup examples. We find that human perceptions do not consistently align with the labels traditionally used for synthetic points, and begin to demonstrate the applicability of these findings to potentially increase the reliability of downstream models, particularly when incorporating human uncertainty. We release all elicited judgments in a new data hub we call H-Mix.
♻ ☆ Reward Poisoning Attacks on Offline Multi-Agent Reinforcement Learning
In offline multi-agent reinforcement learning (MARL), agents estimate policies from a given dataset. We study reward-poisoning attacks in this setting where an exogenous attacker modifies the rewards in the dataset before the agents see the dataset. The attacker wants to guide each agent into a nefarious target policy while minimizing the $L^p$ norm of the reward modification. Unlike attacks on single-agent RL, we show that the attacker can install the target policy as a Markov Perfect Dominant Strategy Equilibrium (MPDSE), which rational agents are guaranteed to follow. This attack can be significantly cheaper than separate single-agent attacks. We show that the attack works on various MARL agents including uncertainty-aware learners, and we exhibit linear programs to efficiently solve the attack problem. We also study the relationship between the structure of the datasets and the minimal attack cost. Our work paves the way for studying defense in offline MARL.
♻ ☆ Vector Optimization with Stochastic Bandit Feedback AISTATS
We introduce vector optimization problems with stochastic bandit feedback, in which preferences among designs are encoded by a polyhedral ordering cone $C$. Our setup generalizes the best arm identification problem to vector-valued rewards by extending the concept of Pareto set beyond multi-objective optimization. We characterize the sample complexity of ($\epsilon,\delta$)-PAC Pareto set identification by defining a new cone-dependent notion of complexity, called the ordering complexity. In particular, we provide gap-dependent and worst-case lower bounds on the sample complexity and show that, in the worst-case, the sample complexity scales with the square of ordering complexity. Furthermore, we investigate the sample complexity of the na\"ive elimination algorithm and prove that it nearly matches the worst-case sample complexity. Finally, we run experiments to verify our theoretical results and illustrate how $C$ and sampling budget affect the Pareto set, the returned ($\epsilon,\delta$)-PAC Pareto set, and the success of identification.
comment: 26 pages, 3 tables, 2 figure; Proceedings of the 26th International Conference on Artificial Intelligence and Statistics (AISTATS) 2023, Valencia, Spain
♻ ☆ Optimal quantum dataset for learning a unitary transformation
Unitary transformations formulate the time evolution of quantum states. How to learn a unitary transformation efficiently is a fundamental problem in quantum machine learning. The most natural and leading strategy is to train a quantum machine learning model based on a quantum dataset. Although the presence of more training data results in better models, using too much data reduces the efficiency of training. In this work, we solve the problem on the minimum size of sufficient quantum datasets for learning a unitary transformation exactly, which reveals the power and limitation of quantum data. First, we prove that the minimum size of a dataset with pure states is $2^n$ for learning an $n$-qubit unitary transformation. To fully explore the capability of quantum data, we introduce a practical quantum dataset consisting of $n+1$ elementary tensor product states that are sufficient for exact training. The main idea is to simplify the structure utilizing decoupling, which leads to an exponential improvement in the size of the datasets with pure states. Furthermore, we show that the size of the quantum dataset with mixed states can be reduced to a constant, which yields an optimal quantum dataset for learning a unitary. We showcase the applications of our results in oracle compiling and Hamiltonian simulation. Notably, to accurately simulate a 3-qubit one-dimensional nearest-neighbor Heisenberg model, our circuit only uses $96$ elementary quantum gates, which is significantly less than $4080$ gates in the circuit constructed by the Trotter-Suzuki product formula.
comment: 11 pages including appendix, v2 added remarks and references, v3 is closed to the published version
♻ ☆ Effects of Parameter Norm Growth During Transformer Training: Inductive Bias from Gradient Descent EMNLP 2021
The capacity of neural networks like the widely adopted transformer is known to be very high. Evidence is emerging that they learn successfully due to inductive bias in the training routine, typically a variant of gradient descent (GD). To better understand this bias, we study the tendency for transformer parameters to grow in magnitude ($\ell_2$ norm) during training, and its implications for the emergent representations within self attention layers. Empirically, we document norm growth in the training of transformer language models, including T5 during its pretraining. As the parameters grow in magnitude, we prove that the network approximates a discretized network with saturated activation functions. Such "saturated" networks are known to have a reduced capacity compared to the full network family that can be described in terms of formal languages and automata. Our results suggest saturation is a new characterization of an inductive bias implicit in GD of particular interest for NLP. We leverage the emergent discrete structure in a saturated transformer to analyze the role of different attention heads, finding that some focus locally on a small number of positions, while other heads compute global averages, allowing counting. We believe understanding the interplay between these two capabilities may shed further light on the structure of computation within large transformers.
comment: Appeared at EMNLP 2021. March 7, 2023: Removed irreproducible numbers reported in a footnote with erratum note
♻ ☆ Explainability in Deep Reinforcement Learning, a Review into Current Methods and Applications
The use of Deep Reinforcement Learning (DRL) schemes has increased dramatically since their first introduction in 2015. Though uses in many different applications are being found, they still have a problem with the lack of interpretability. This has bread a lack of understanding and trust in the use of DRL solutions from researchers and the general public. To solve this problem, the field of Explainable Artificial Intelligence (XAI) has emerged. This entails a variety of different methods that look to open the DRL black boxes, ranging from the use of interpretable symbolic Decision Trees (DT) to numerical methods like Shapley Values. This review looks at which methods are being used and for which applications. This is done to identify which models are the best suited to each application or if a method is being underutilised.
comment: 30 pages, 6 figures, Paper Review
♻ ☆ Task-Adaptive Meta-Learning Framework for Advancing Spatial Generalizability AAAI
Spatio-temporal machine learning is critically needed for a variety of societal applications, such as agricultural monitoring, hydrological forecast, and traffic management. These applications greatly rely on regional features that characterize spatial and temporal differences. However, spatio-temporal data often exhibit complex patterns and significant data variability across different locations. The labels in many real-world applications can also be limited, which makes it difficult to separately train independent models for different locations. Although meta learning has shown promise in model adaptation with small samples, existing meta learning methods remain limited in handling a large number of heterogeneous tasks, e.g., a large number of locations with varying data patterns. To bridge the gap, we propose task-adaptive formulations and a model-agnostic meta-learning framework that ensembles regionally heterogeneous data into location-sensitive meta tasks. We conduct task adaptation following an easy-to-hard task hierarchy in which different meta models are adapted to tasks of different difficulty levels. One major advantage of our proposed method is that it improves the model adaptation to a large number of heterogeneous tasks. It also enhances the model generalization by automatically adapting the meta model of the corresponding difficulty level to any new tasks. We demonstrate the superiority of our proposed framework over a diverse set of baselines and state-of-the-art meta-learning frameworks. Our extensive experiments on real crop yield data show the effectiveness of the proposed method in handling spatial-related heterogeneous tasks in real societal applications.
comment: In the Thirty-Seventh AAAI Conference on Artificial Intelligence, February 2023
♻ ☆ Py-Feat: Python Facial Expression Analysis Toolbox
Studying facial expressions is a notoriously difficult endeavor. Recent advances in the field of affective computing have yielded impressive progress in automatically detecting facial expressions from pictures and videos. However, much of this work has yet to be widely disseminated in social science domains such as psychology. Current state of the art models require considerable domain expertise that is not traditionally incorporated into social science training programs. Furthermore, there is a notable absence of user-friendly and open-source software that provides a comprehensive set of tools and functions that support facial expression research. In this paper, we introduce Py-Feat, an open-source Python toolbox that provides support for detecting, preprocessing, analyzing, and visualizing facial expression data. Py-Feat makes it easy for domain experts to disseminate and benchmark computer vision models and also for end users to quickly process, analyze, and visualize face expression data. We hope this platform will facilitate increased use of facial expression data in human behavior research.
♻ ☆ Estimating a Brain Network Predictive of Stress and Genotype with Supervised Autoencoders
Targeted stimulation of the brain has the potential to treat mental illnesses. We propose an approach to help design the stimulation protocol by identifying electrical dynamics across many brain regions that relate to illness states. We model multi-region electrical activity as a superposition of activity from latent networks, where the weights on the latent networks relate to an outcome of interest. In order to improve on drawbacks of latent factor modeling in this context, we focus on supervised autoencoders (SAEs), which can improve predictive performance while maintaining a generative model. We explain why SAEs yield improved predictions, describe the distributional assumptions under which SAEs are an appropriate modeling choice, and provide modeling constraints to ensure biological relevance of the learned network. We use the analysis strategy to find a network associated with stress that characterizes a genotype associated with bipolar disorder. This discovered network aligns with a previously used stimulation technique, providing experimental validation of our approach.
comment: 43 pages, 9 figures
♻ ☆ The Open Catalyst 2022 (OC22) Dataset and Challenges for Oxide Electrocatalysts
The development of machine learning models for electrocatalysts requires a broad set of training data to enable their use across a wide variety of materials. One class of materials that currently lacks sufficient training data is oxides, which are critical for the development of OER catalysts. To address this, we developed the OC22 dataset, consisting of 62,331 DFT relaxations (~9,854,504 single point calculations) across a range of oxide materials, coverages, and adsorbates. We define generalized total energy tasks that enable property prediction beyond adsorption energies; we test baseline performance of several graph neural networks; and we provide pre-defined dataset splits to establish clear benchmarks for future efforts. In the most general task, GemNet-OC sees a ~36% improvement in energy predictions when combining the chemically dissimilar OC20 and OC22 datasets via fine-tuning. Similarly, we achieved a ~19% improvement in total energy predictions on OC20 and a ~9% improvement in force predictions in OC22 when using joint training. We demonstrate the practical utility of a top performing model by capturing literature adsorption energies and important OER scaling relationships. We expect OC22 to provide an important benchmark for models seeking to incorporate intricate long-range electrostatic and magnetic interactions in oxide surfaces. Dataset and baseline models are open sourced, and a public leaderboard is available to encourage continued community developments on the total energy tasks and data.
comment: 50 pages, 14 figures
♻ ☆ Neural Collapse with Normalized Features: A Geometric Analysis over the Riemannian Manifold NeurIPS'22
When training overparameterized deep networks for classification tasks, it has been widely observed that the learned features exhibit a so-called "neural collapse" phenomenon. More specifically, for the output features of the penultimate layer, for each class the within-class features converge to their means, and the means of different classes exhibit a certain tight frame structure, which is also aligned with the last layer's classifier. As feature normalization in the last layer becomes a common practice in modern representation learning, in this work we theoretically justify the neural collapse phenomenon for normalized features. Based on an unconstrained feature model, we simplify the empirical loss function in a multi-class classification task into a nonconvex optimization problem over the Riemannian manifold by constraining all features and classifiers over the sphere. In this context, we analyze the nonconvex landscape of the Riemannian optimization problem over the product of spheres, showing a benign global landscape in the sense that the only global minimizers are the neural collapse solutions while all other critical points are strict saddles with negative curvature. Experimental results on practical deep networks corroborate our theory and demonstrate that better representations can be learned faster via feature normalization.
comment: The first two authors contributed to this work equally; 38 pages, 13 figures. Accepted at NeurIPS'22
♻ ☆ On the Generalization Power of Overfitted Two-Layer Neural Tangent Kernel Models ICML21
In this paper, we study the generalization performance of min $\ell_2$-norm overfitting solutions for the neural tangent kernel (NTK) model of a two-layer neural network with ReLU activation that has no bias term. We show that, depending on the ground-truth function, the test error of overfitted NTK models exhibits characteristics that are different from the "double-descent" of other overparameterized linear models with simple Fourier or Gaussian features. Specifically, for a class of learnable functions, we provide a new upper bound of the generalization error that approaches a small limiting value, even when the number of neurons $p$ approaches infinity. This limiting value further decreases with the number of training samples $n$. For functions outside of this class, we provide a lower bound on the generalization error that does not diminish to zero even when $n$ and $p$ are both large.
comment: Published in ICML21. This version fixes an error of Lemma 31 and other parts affected by this error. The main results remain the same except some small changes on certain coefficients of Eq.(9)
♻ ☆ Causal Representation Learning for Instantaneous and Temporal Effects in Interactive Systems ICLR
Causal representation learning is the task of identifying the underlying causal variables and their relations from high-dimensional observations, such as images. Recent work has shown that one can reconstruct the causal variables from temporal sequences of observations under the assumption that there are no instantaneous causal relations between them. In practical applications, however, our measurement or frame rate might be slower than many of the causal effects. This effectively creates "instantaneous" effects and invalidates previous identifiability results. To address this issue, we propose iCITRIS, a causal representation learning method that allows for instantaneous effects in intervened temporal sequences when intervention targets can be observed, e.g., as actions of an agent. iCITRIS identifies the potentially multidimensional causal variables from temporal observations, while simultaneously using a differentiable causal discovery method to learn their causal graph. In experiments on three datasets of interactive systems, iCITRIS accurately identifies the causal variables and their causal graph.
comment: Published at International Conference on Learning Representations (ICLR), 2023
Multimedia
☆ BIRD-PCC: Bi-directional Range Image-based Deep LiDAR Point Cloud Compression ICASSP 2023
The large amount of data collected by LiDAR sensors brings the issue of LiDAR point cloud compression (PCC). Previous works on range image-based LiDAR PCC follow the predictive coding paradigm, structuring a simple prototype of a coding framework. However, their prediction methods give an inaccurate result due to the negligence of invalid pixels in range images and the omission of future frames in the time step. Moreover, their handcrafted design of residual coding methods could not fully exploit spatial redundancy. To remedy this, we propose a coding framework BIRD-PCC. Our prediction module is aware of the coordinates of invalid pixels in range images and takes a bidirectional scheme. Also, we introduce a deep-learned residual coding module that can further exploit spatial redundancy within a residual frame. Experiments conducted on SemanticKITTI and KITTI-360 datasets show that BIRD-PCC outperforms other methods in most bitrate conditions and generalizes well to unseen environments.
comment: Accepted to ICASSP 2023
☆ Leveraging Pre-trained AudioLDM for Sound Generation: A Benchmark Study
Deep neural networks have recently achieved breakthroughs in sound generation. Despite the outstanding sample quality, current sound generation models face issues on small-scale datasets (e.g., overfitting and low coverage of sound classes), significantly limiting performance. In this paper, we make the first attempt to investigate the benefits of pre-training on sound generation with AudioLDM, the cutting-edge model for audio generation, as the backbone. Our study demonstrates the advantages of the pre-trained AudioLDM, especially in data-scarcity scenarios. In addition, the baselines and evaluation protocol for sound generation systems are not consistent enough to compare different studies directly. Aiming to facilitate further study on sound generation tasks, we benchmark the sound generation task on various frequently-used datasets. We hope our results on transfer learning and benchmarks can provide references for further research on conditional sound generation.
comment: EUSIPCO 2023
☆ FSVVD: A Dataset of Full Scene Volumetric Video
Recent years have witnessed a rapid development of immersive multimedia which bridges the gap between the real world and virtual space. Volumetric videos, as an emerging representative 3D video paradigm that empowers extended reality, stand out to provide unprecedented immersive and interactive video watching experience. Despite the tremendous potential, the research towards 3D volumetric video is still in its infancy, relying on sufficient and complete datasets for further exploration. However, existing related volumetric video datasets mostly only include a single object, lacking details about the scene and the interaction between them. In this paper, we focus on the current most widely used data format, point cloud, and for the first time release a full-scene volumetric video dataset that includes multiple people and their daily activities interacting with the external environments. Comprehensive dataset description and analysis are conducted, with potential usage of this dataset. The dataset and additional tools can be accessed via the following website: https://cuhksz-inml.github.io/full_scene_volumetric_video_dataset/.
comment: Accepted by MMSys'23 Open Dataset and Software Track, A preliminary version. The dataset and additional tools can be accessed via https://cuhksz-inml.github.io/full_scene_volumetric_video_dataset/
☆ Approach to Learning Generalized Audio Representation Through Batch Embedding Covariance Regularization and Constant-Q Transforms
General-purpose embedding is highly desirable for few-shot even zero-shot learning in many application scenarios, including audio tasks. In order to understand representations better, we conducted a thorough error analysis and visualization of HEAR 2021 submission results. Inspired by the analysis, this work experiments with different front-end audio preprocessing methods, including Constant-Q Transform (CQT) and Short-time Fourier transform (STFT), and proposes a Batch Embedding Covariance Regularization (BECR) term to uncover a more holistic simulation of the frequency information received by the human auditory system. We tested the models on the suite of HEAR 2021 tasks, which encompass a broad category of tasks. Preliminary results show (1) the proposed BECR can incur a more dispersed embedding on the test set, (2) BECR improves the PaSST model without extra computation complexity, and (3) STFT preprocessing outperforms CQT in all tasks we tested. Github:https://github.com/ankitshah009/general_audio_embedding_hear_2021
comment: Technical report, 10 pages
♻ ☆ Compose & Embellish: Well-Structured Piano Performance Generation via A Two-Stage Approach ICASSP
Even with strong sequence models like Transformers, generating expressive piano performances with long-range musical structures remains challenging. Meanwhile, methods to compose well-structured melodies or lead sheets (melody + chords), i.e., simpler forms of music, gained more success. Observing the above, we devise a two-stage Transformer-based framework that Composes a lead sheet first, and then Embellishes it with accompaniment and expressive touches. Such a factorization also enables pretraining on non-piano data. Our objective and subjective experiments show that Compose & Embellish shrinks the gap in structureness between a current state of the art and real performances by half, and improves other musical aspects such as richness and coherence as well.
comment: Accepted to International Conference on Acoustics, Speech, and Signal Processing (ICASSP) 2023
♻ ☆ Deep Learning for Predictive Analytics in Reversible Steganography
Deep learning is regarded as a promising solution for reversible steganography. There is an accelerating trend of representing a reversible steo-system by monolithic neural networks, which bypass intermediate operations in traditional pipelines of reversible steganography. This end-to-end paradigm, however, suffers from imperfect reversibility. By contrast, the modular paradigm that incorporates neural networks into modules of traditional pipelines can stably guarantee reversibility with mathematical explainability. Prediction-error modulation is a well-established reversible steganography pipeline for digital images. It consists of a predictive analytics module and a reversible coding module. Given that reversibility is governed independently by the coding module, we narrow our focus to the incorporation of neural networks into the analytics module, which serves the purpose of predicting pixel intensities and a pivotal role in determining capacity and imperceptibility. The objective of this study is to evaluate the impacts of different training configurations upon predictive accuracy of neural networks and provide practical insights. In particular, we investigate how different initialisation strategies for input images may affect the learning process and how different training strategies for dual-layer prediction respond to the problem of distributional shift. Furthermore, we compare steganographic performance of various model architectures with different loss functions.
♻ ☆ Bayesian Neural Networks for Reversible Steganography
Recent advances in deep learning have led to a paradigm shift in the field of reversible steganography. A fundamental pillar of reversible steganography is predictive modelling which can be realised via deep neural networks. However, non-trivial errors exist in inferences about some out-of-distribution and noisy data. In view of this issue, we propose to consider uncertainty in predictive models based upon a theoretical framework of Bayesian deep learning, thereby creating an adaptive steganographic system. Most modern deep-learning models are regarded as deterministic because they only offer predictions while failing to provide uncertainty measurement. Bayesian neural networks bring a probabilistic perspective to deep learning and can be regarded as self-aware intelligent machinery; that is, a machine that knows its own limitations. To quantify uncertainty, we apply Bayesian statistics to model the predictive distribution and approximate it through Monte Carlo sampling with stochastic forward passes. We further show that predictive uncertainty can be disentangled into aleatoric and epistemic uncertainties and these quantities can be learnt unsupervised. Experimental results demonstrate an improvement delivered by Bayesian uncertainty analysis upon steganographic rate-distortion performance.
♻ ☆ On the predictability in reversible steganography
Artificial neural networks have advanced the frontiers of reversible steganography. The core strength of neural networks is the ability to render accurate predictions for a bewildering variety of data. Residual modulation is recognised as the most advanced reversible steganographic algorithm for digital images. The pivot of this algorithm is predictive analytics in which pixel intensities are predicted given some pixel-wise contextual information. This task can be perceived as a low-level vision problem and hence neural networks for addressing a similar class of problems can be deployed. On top of the prior art, this paper investigates predictability of pixel intensities based on supervised and unsupervised learning frameworks. Predictability analysis enables adaptive data embedding, which in turn leads to a better trade-off between capacity and imperceptibility. While conventional methods estimate predictability by the statistics of local image patterns, learning-based frameworks consider further the degree to which correct predictions can be made by a designated predictor. Not only should the image patterns be taken into account but also the predictor in use. Experimental results show that steganographic performance can be significantly improved by incorporating the learning-based predictability analysers into a reversible steganographic system.
♻ ☆ Exploring Video Quality Assessment on User Generated Contents from Aesthetic and Technical Perspectives
The rapid increase in user-generated-content (UGC) videos calls for the development of effective video quality assessment (VQA) algorithms. However, the objective of the UGC-VQA problem is still ambiguous and can be viewed from two perspectives: the technical perspective, measuring the perception of distortions; and the aesthetic perspective, which relates to preference and recommendation on contents. To understand how these two perspectives affect overall subjective opinions in UGC-VQA, we conduct a large-scale subjective study to collect human quality opinions on overall quality of videos as well as perceptions from aesthetic and technical perspectives. The collected Disentangled Video Quality Database (DIVIDE-3k) confirms that human quality opinions on UGC videos are universally and inevitably affected by both aesthetic and technical perspectives. In light of this, we propose the Disentangled Objective Video Quality Evaluator (DOVER) to learn the quality of UGC videos based on the two perspectives. The DOVER proves state-of-the-art performance in UGC-VQA under very high efficiency. With perspective opinions in DIVIDE-3k, we further propose DOVER++, the first approach to provide reliable clear-cut quality evaluations from a single aesthetic or technical perspective. Code at https://github.com/VQAssessment/DOVER.
♻ ☆ Action-GPT: Leveraging Large-scale Language Models for Improved and Generalized Action Generation
We introduce Action-GPT, a plug-and-play framework for incorporating Large Language Models (LLMs) into text-based action generation models. Action phrases in current motion capture datasets contain minimal and to-the-point information. By carefully crafting prompts for LLMs, we generate richer and fine-grained descriptions of the action. We show that utilizing these detailed descriptions instead of the original action phrases leads to better alignment of text and motion spaces. We introduce a generic approach compatible with stochastic (e.g. VAE-based) and deterministic (e.g. MotionCLIP) text-to-motion models. In addition, the approach enables multiple text descriptions to be utilized. Our experiments show (i) noticeable qualitative and quantitative improvement in the quality of synthesized motions, (ii) benefits of utilizing multiple LLM-generated descriptions, (iii) suitability of the prompt function, and (iv) zero-shot generation capabilities of the proposed approach. Project page: https://actiongpt.github.io
comment: Code, pretrained models and sample videos will be made available at \url{https://actiongpt.github.io}
♻ ☆ Confidence-based Event-centric Online Video Question Answering on a Newly Constructed ATBS Dataset ICASSP 2023
Deep neural networks facilitate video question answering (VideoQA), but the real-world applications on video streams such as CCTV and live cast place higher demands on the solver. To address the challenges of VideoQA on long videos of unknown length, we define a new set of problems called Online Open-ended Video Question Answering (O^2VQA). It requires an online state-updating mechanism for the solver to decide if the collected information is sufficient to conclude an answer. We then propose a Confidence-based Event-centric Online Video Question Answering (CEO-VQA) model to solve this problem. Furthermore, a dataset called Answer Target in Background Stream (ATBS) is constructed to evaluate this newly developed online VideoQA application. Compared to the baseline VideoQA method that watches the whole video, the experimental results show that the proposed method achieves a significant performance gain.
comment: Accepted for publication at the 2023 IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2023)
♻ ☆ Time-frequency Network for Robust Speaker Recognition
The wide deployment of speech-based biometric systems usually demands high-performance speaker recognition algorithms. However, most of the prior works for speaker recognition either process the speech in the frequency domain or time domain, which may produce suboptimal results because both time and frequency domains are important for speaker recognition. In this paper, we attempt to analyze the speech signal in both time and frequency domains and propose the time-frequency network~(TFN) for speaker recognition by extracting and fusing the features in the two domains. Based on the recent advance of deep neural networks, we propose a convolution neural network to encode the raw speech waveform and the frequency spectrum into domain-specific features, which are then fused and transformed into a classification feature space for speaker recognition. Experimental results on the publicly available datasets TIMIT and LibriSpeech show that our framework is effective to combine the information in the two domains and performs better than the state-of-the-art methods for speaker recognition.
comment: 5pages, 3 figures
Computation and Language
☆ Enhancing Activity Prediction Models in Drug Discovery with the Ability to Understand Human Language
Activity and property prediction models are the central workhorses in drug discovery and materials sciences, but currently they have to be trained or fine-tuned for new tasks. Without training or fine-tuning, scientific language models could be used for such low-data tasks through their announced zero- and few-shot capabilities. However, their predictive quality at activity prediction is lacking. In this work, we envision a novel type of activity prediction model that is able to adapt to new prediction tasks at inference time, via understanding textual information describing the task. To this end, we propose a new architecture with separate modules for chemical and natural language inputs, and a contrastive pre-training objective on data from large biochemical databases. In extensive experiments, we show that our method CLAMP yields improved predictive performance on few-shot learning benchmarks and zero-shot problems in drug discovery. We attribute the advances of our method to the modularized architecture and to our pre-training objective.
comment: 15 pages + 18 pages appendix
☆ AmQA: Amharic Question Answering Dataset
Question Answering (QA) returns concise answers or answer lists from natural language text given a context document. Many resources go into curating QA datasets to advance robust models' development. There is a surge of QA datasets for languages like English, however, this is not true for Amharic. Amharic, the official language of Ethiopia, is the second most spoken Semitic language in the world. There is no published or publicly available Amharic QA dataset. Hence, to foster the research in Amharic QA, we present the first Amharic QA (AmQA) dataset. We crowdsourced 2628 question-answer pairs over 378 Wikipedia articles. Additionally, we run an XLMR Large-based baseline model to spark open-domain QA research interest. The best-performing baseline achieves an F-score of 69.58 and 71.74 in reader-retriever QA and reading comprehension settings respectively.
☆ The AI Ghostwriter Effect: Users Do Not Perceive Ownership of AI-Generated Text But Self-Declare as Authors
Human-AI interaction in text production increases complexity in authorship. In two empirical studies (n1 = 30 & n2 = 96), we investigate authorship and ownership in human-AI collaboration for personalized language generation models. We show an AI Ghostwriter Effect: Users do not consider themselves the owners and authors of AI-generated text but refrain from publicly declaring AI authorship. The degree of personalization did not impact the AI Ghostwriter Effect, and control over the model increased participants' sense of ownership. We also found that the discrepancy between the sense of ownership and the authorship declaration is stronger in interactions with a human ghostwriter and that people use similar rationalizations for authorship in AI ghostwriters and human ghostwriters. We discuss how our findings relate to psychological ownership and human-AI interaction to lay the foundations for adapting authorship frameworks and user interfaces in AI in text-generation tasks.
comment: Pre-print; currently under review
☆ Faithfulness-Aware Decoding Strategies for Abstractive Summarization EACL 2023
Despite significant progress in understanding and improving faithfulness in abstractive summarization, the question of how decoding strategies affect faithfulness is less studied. We present a systematic study of the effect of generation techniques such as beam search and nucleus sampling on faithfulness in abstractive summarization. We find a consistent trend where beam search with large beam sizes produces the most faithful summaries while nucleus sampling generates the least faithful ones. We propose two faithfulness-aware generation methods to further improve faithfulness over current generation techniques: (1) ranking candidates generated by beam search using automatic faithfulness metrics and (2) incorporating lookahead heuristics that produce a faithfulness score on the future summary. We show that both generation methods significantly improve faithfulness across two datasets as evaluated by four automatic faithfulness metrics and human evaluation. To reduce computational cost, we demonstrate a simple distillation approach that allows the model to generate faithful summaries with just greedy decoding. Our code is publicly available at https://github.com/amazon-science/faithful-summarization-generation
comment: EACL 2023 (17 pages)
☆ Choice Over Control: How Users Write with Large Language Models using Diegetic and Non-Diegetic Prompting
We propose a conceptual perspective on prompts for Large Language Models (LLMs) that distinguishes between (1) diegetic prompts (part of the narrative, e.g. "Once upon a time, I saw a fox..."), and (2) non-diegetic prompts (external, e.g. "Write about the adventures of the fox."). With this lens, we study how 129 crowd workers on Prolific write short texts with different user interfaces (1 vs 3 suggestions, with/out non-diegetic prompts; implemented with GPT-3): When the interface offered multiple suggestions and provided an option for non-diegetic prompting, participants preferred choosing from multiple suggestions over controlling them via non-diegetic prompts. When participants provided non-diegetic prompts it was to ask for inspiration, topics or facts. Single suggestions in particular were guided both with diegetic and non-diegetic information. This work informs human-AI interaction with generative models by revealing that (1) writing non-diegetic prompts requires effort, (2) people combine diegetic and non-diegetic prompting, and (3) they use their draft (i.e. diegetic information) and suggestion timing to strategically guide LLMs.
comment: 17 pages, 9 figures, 3 tables, ACM CHI 2023
☆ Neighborhood Contrastive Transformer for Change Captioning
Change captioning is to describe the semantic change between a pair of similar images in natural language. It is more challenging than general image captioning, because it requires capturing fine-grained change information while being immune to irrelevant viewpoint changes, and solving syntax ambiguity in change descriptions. In this paper, we propose a neighborhood contrastive transformer to improve the model's perceiving ability for various changes under different scenes and cognition ability for complex syntax structure. Concretely, we first design a neighboring feature aggregating to integrate neighboring context into each feature, which helps quickly locate the inconspicuous changes under the guidance of conspicuous referents. Then, we devise a common feature distilling to compare two images at neighborhood level and extract common properties from each image, so as to learn effective contrastive information between them. Finally, we introduce the explicit dependencies between words to calibrate the transformer decoder, which helps better understand complex syntax structure during training. Extensive experimental results demonstrate that the proposed method achieves the state-of-the-art performance on three public datasets with different change scenarios. The code is available at https://github.com/tuyunbin/NCT.
comment: Accepted by IEEE TMM
☆ IPA-CLIP: Integrating Phonetic Priors into Vision and Language Pretraining
Recently, large-scale Vision and Language (V\&L) pretraining has become the standard backbone of many multimedia systems. While it has shown remarkable performance even in unseen situations, it often performs in ways not intuitive to humans. Particularly, they usually do not consider the pronunciation of the input, which humans would utilize to understand language, especially when it comes to unknown words. Thus, this paper inserts phonetic prior into Contrastive Language-Image Pretraining (CLIP), one of the V\&L pretrained models, to make it consider the pronunciation similarity among its pronunciation inputs. To achieve this, we first propose a phoneme embedding that utilizes the phoneme relationships provided by the International Phonetic Alphabet (IPA) chart as a phonetic prior. Next, by distilling the frozen CLIP text encoder, we train a pronunciation encoder employing the IPA-based embedding. The proposed model named IPA-CLIP comprises this pronunciation encoder and the original CLIP encoders (image and text). Quantitative evaluation reveals that the phoneme distribution on the embedding space represents phonetic relationships more accurately when using the proposed phoneme embedding. Furthermore, in some multimodal retrieval tasks, we confirm that the proposed pronunciation encoder enhances the performance of the text encoder and that the pronunciation encoder handles nonsense words in a more phonetic manner than the text encoder. Finally, qualitative evaluation verifies the correlation between the pronunciation encoder and human perception regarding pronunciation similarity.
comment: 11 pages, 8 figures, 5 Tables
☆ IFAN: An Explainability-Focused Interaction Framework for Humans and NLP Models ACL
Interpretability and human oversight are fundamental pillars of deploying complex NLP models into real-world applications. However, applying explainability and human-in-the-loop methods requires technical proficiency. Despite existing toolkits for model understanding and analysis, options to integrate human feedback are still limited. We propose IFAN, a framework for real-time explanation-based interaction with NLP models. Through IFAN's interface, users can provide feedback to selected model explanations, which is then integrated through adapter layers to align the model with human rationale. We show the system to be effective in debiasing a hate speech classifier with minimal performance loss. IFAN also offers a visual admin system and API to manage models (and datasets) as well as control access rights. A demo is live at https://ifan.ml/
comment: ACL Demo 2023 Submission
☆ Towards Zero-Shot Functional Compositionality of Language Models
Large Pre-trained Language Models (PLM) have become the most desirable starting point in the field of NLP, as they have become remarkably good at solving many individual tasks. Despite such success, in this paper, we argue that current paradigms of working with PLMs are neglecting a critical aspect of modeling human intelligence: functional compositionality. Functional compositionality - the ability to compose learned tasks - has been a long-standing challenge in the field of AI (and many other fields) as it is considered one of the hallmarks of human intelligence. An illustrative example of such is cross-lingual summarization, where a bilingual person (English-French) could directly summarize an English document into French sentences without having to translate the English document or summary into French explicitly. We discuss why this matter is an important open problem that requires further attention from the field. Then, we show that current PLMs (e.g., GPT-2 and T5) don't have functional compositionality yet and it is far from human-level generalizability. Finally, we suggest several research directions that could push the field towards zero-shot functional compositionality of language models.
☆ Crowdsourcing on Sensitive Data with Privacy-Preserving Text Rewriting
Most tasks in NLP require labeled data. Data labeling is often done on crowdsourcing platforms due to scalability reasons. However, publishing data on public platforms can only be done if no privacy-relevant information is included. Textual data often contains sensitive information like person names or locations. In this work, we investigate how removing personally identifiable information (PII) as well as applying differential privacy (DP) rewriting can enable text with privacy-relevant information to be used for crowdsourcing. We find that DP-rewriting before crowdsourcing can preserve privacy while still leading to good label quality for certain tasks and data. PII-removal led to good label quality in all examined tasks, however, there are no privacy guarantees given.
☆ DeCap: Decoding CLIP Latents for Zero-Shot Captioning via Text-Only Training ICLR 2023
Large-scale pre-trained multi-modal models (e.g., CLIP) demonstrate strong zero-shot transfer capability in many discriminative tasks. Their adaptation to zero-shot image-conditioned text generation tasks has drawn increasing interest. Prior arts approach to zero-shot captioning by either utilizing the existing large language models (e.g., GPT-2) or pre-training the encoder-decoder network in an end-to-end manner. In this work, we propose a simple framework, named DeCap, for zero-shot captioning. We introduce a lightweight visual-aware language decoder. This decoder is both data-efficient and computation-efficient: 1) it only requires the text data for training, easing the burden on the collection of paired data. 2) it does not require end-to-end training. When trained with text-only data, the decoder takes the text embedding extracted from the off-the-shelf CLIP encoder as a prefix embedding. The challenge is that the decoder is trained on the text corpus but at the inference stage, it needs to generate captions based on visual inputs. The modality gap issue is widely observed in multi-modal contrastive models that prevents us from directly taking the visual embedding as the prefix embedding. We propose a training-free mechanism to reduce the modality gap. We project the visual embedding into the CLIP text embedding space, while the projected embedding retains the information of the visual input. Taking the projected embedding as the prefix embedding, the decoder generates high-quality descriptions that match the visual input. The experiments show that DeCap outperforms other zero-shot captioning methods and unpaired captioning methods on the typical image captioning benchmarks, i.e., MSCOCO and NoCaps.
comment: Accepted by ICLR 2023. Code is available at https://github.com/dhg-wei/DeCap
☆ NxPlain: Web-based Tool for Discovery of Latent Concepts EACL 2023
The proliferation of deep neural networks in various domains has seen an increased need for the interpretability of these models, especially in scenarios where fairness and trust are as important as model performance. A lot of independent work is being carried out to: i) analyze what linguistic and non-linguistic knowledge is learned within these models, and ii) highlight the salient parts of the input. We present NxPlain, a web application that provides an explanation of a model's prediction using latent concepts. NxPlain discovers latent concepts learned in a deep NLP model, provides an interpretation of the knowledge learned in the model, and explains its predictions based on the used concepts. The application allows users to browse through the latent concepts in an intuitive order, letting them efficiently scan through the most salient concepts with a global corpus level view and a local sentence-level view. Our tool is useful for debugging, unraveling model bias, and for highlighting spurious correlations in a model. A hosted demo is available here: https://nxplain.qcri.org.
comment: EACL 2023
☆ xCodeEval: A Large Scale Multilingual Multitask Benchmark for Code Understanding, Generation, Translation and Retrieval
The ability to solve problems is a hallmark of intelligence and has been an enduring goal in AI. AI systems that can create programs as solutions to problems or assist developers in writing programs can increase productivity and make programming more accessible. Recently, pre-trained large language models have shown impressive abilities in generating new codes from natural language descriptions, repairing buggy codes, translating codes between languages, and retrieving relevant code segments. However, the evaluation of these models has often been performed in a scattered way on only one or two specific tasks, in a few languages, at a partial granularity (e.g., function) level and in many cases without proper training data. Even more concerning is that in most cases the evaluation of generated codes has been done in terms of mere lexical overlap rather than actual execution whereas semantic similarity (or equivalence) of two code segments depends only on their ``execution similarity'', i.e., being able to get the same output for a given input.
☆ HiCLIP: Contrastive Language-Image Pretraining with Hierarchy-aware Attention ICLR 2023
The success of large-scale contrastive vision-language pretraining (CLIP) has benefited both visual recognition and multimodal content understanding. The concise design brings CLIP the advantage in inference efficiency against other vision-language models with heavier cross-attention fusion layers, making it a popular choice for a wide spectrum of downstream tasks. However, CLIP does not explicitly capture the hierarchical nature of high-level and fine-grained semantics conveyed in images and texts, which is arguably critical to vision-language understanding and reasoning. To this end, we equip both the visual and language branches in CLIP with hierarchy-aware attentions, namely Hierarchy-aware CLIP (HiCLIP), to progressively discover semantic hierarchies layer-by-layer from both images and texts in an unsupervised manner. As a result, such hierarchical aggregation significantly improves the cross-modal alignment. To demonstrate the advantages of HiCLIP, we conduct qualitative analysis on its unsupervised hierarchy induction during inference, as well as extensive quantitative experiments on both visual recognition and vision-language downstream tasks.
comment: Accepted at ICLR 2023
☆ GlobalNER: Incorporating Non-local Information into Named Entity Recognition
Nowadays, many Natural Language Processing (NLP) tasks see the demand for incorporating knowledge external to the local information to further improve the performance. However, there is little related work on Named Entity Recognition (NER), which is one of the foundations of NLP. Specifically, no studies were conducted on the query generation and re-ranking for retrieving the related information for the purpose of improving NER. This work demonstrates the effectiveness of a DNN-based query generation method and a mention-aware re-ranking architecture based on BERTScore particularly for NER. In the end, a state-of-the-art performance of 61.56 micro-f1 score on WNUT17 dataset is achieved.
comment: 13 pages, 5 figures
☆ OpenICL: An Open-Source Framework for In-context Learning
In recent years, In-context Learning (ICL) has gained increasing attention and emerged as the new paradigm for large language model (LLM) evaluation. Unlike traditional fine-tuning methods, ICL instead adapts the pre-trained models to unseen tasks without any parameter updates. However, the implementation of ICL is sophisticated due to the diverse retrieval and inference methods involved, as well as the varying pre-processing requirements for different models, datasets, and tasks. A unified and flexible framework for ICL is urgently needed to ease the implementation of the aforementioned components. To facilitate ICL research, we introduce OpenICL, an open-source toolkit for ICL and LLM evaluation. OpenICL is research-friendly with a highly flexible architecture that users can easily combine different components to suit their needs. It also provides various state-of-the-art retrieval and inference methods to streamline the process of adapting ICL to cutting-edge research. The effectiveness of OpenICL has been validated on a wide range of NLP tasks, including classification, QA, machine translation, and semantic parsing. As a side-product, we found OpenICL to be an efficient yet robust tool for LLMs evaluation. OpenICL is released at https://github.com/Shark-NLP/OpenICL
☆ Dynamic Prompting: A Unified Framework for Prompt Tuning
It has been demonstrated that prompt tuning is highly effective in efficiently eliciting knowledge from language models (LMs). However, the prompt tuning still lags behind fine-tuning, especially when the LMs are small. P-tuning v2 (Liu et al., 2021b) makes it comparable with finetuning by adding continuous prompts for every layer of the pre-trained model. However, prepending fixed soft prompts for all instances, regardless of their discrepancy, is doubtful. In particular, the inserted prompt position, length, and the representations of prompts for diversified instances through different tasks could all affect the prompt tuning performance. To fill this gap, we propose dynamic prompting (DP): the position, length, and prompt representation can all be dynamically optimized with respect to different tasks and instances. We conduct comprehensive experiments on the SuperGlue benchmark to validate our hypothesis and demonstrate substantial improvements. We also derive a unified framework for supporting our dynamic prompting strategy. In particular, we use a simple learning network and Gumble- Softmax for learning instance-dependent guidance. Experimental results show that simple instance-level position-aware soft prompts can improve the classification accuracy of up to 6 points on average on five datasets, reducing its gap with fine-tuning. Besides, we also prove its universal usefulness under full-data, few-shot, and multitask regimes. Combining them together can even further unleash the power of DP, narrowing the distance between finetuning.
comment: Work in progress
☆ Multitask Prompt Tuning Enables Parameter-Efficient Transfer Learning ICLR 2023
Prompt tuning, in which a base pretrained model is adapted to each task via conditioning on learned prompt vectors, has emerged as a promising approach for efficiently adapting large language models to multiple downstream tasks. However, existing methods typically learn soft prompt vectors from scratch, and it has not been clear how to exploit the rich cross-task knowledge with prompt vectors in a multitask learning setting. We propose multitask prompt tuning (MPT), which first learns a single transferable prompt by distilling knowledge from multiple task-specific source prompts. We then learn multiplicative low rank updates to this shared prompt to efficiently adapt it to each downstream target task. Extensive experiments on 23 NLP datasets demonstrate that our proposed approach outperforms the state-of-the-art methods, including the full finetuning baseline in some cases, despite only tuning 0.035% as many task-specific parameters.
comment: ICLR 2023. Project page: https://zhenwang9102.github.io/mpt.html
☆ A Multi-Grained Self-Interpretable Symbolic-Neural Model For Single/Multi-Labeled Text Classification ICLR 2023
Deep neural networks based on layer-stacking architectures have historically suffered from poor inherent interpretability. Meanwhile, symbolic probabilistic models function with clear interpretability, but how to combine them with neural networks to enhance their performance remains to be explored. In this paper, we try to marry these two systems for text classification via a structured language model. We propose a Symbolic-Neural model that can learn to explicitly predict class labels of text spans from a constituency tree without requiring any access to span-level gold labels. As the structured language model learns to predict constituency trees in a self-supervised manner, only raw texts and sentence-level labels are required as training data, which makes it essentially a general constituent-level self-interpretable classification model. Our experiments demonstrate that our approach could achieve good prediction accuracy in downstream tasks. Meanwhile, the predicted span labels are consistent with human rationales to a certain degree.
comment: ICLR 2023
☆ Reducing Spurious Correlations for Aspect-Based Sentiment Analysis with Variational Information Bottleneck and Contrastive Learning
The literature on aspect-based sentiment analysis (ABSA) has been overwhelmed by deep neural networks, yielding state-of-the-art results for ABSA. However, these deep models are susceptible to learning spurious correlations between input features and output labels, which in general suffer from poor robustness and generalization. In this paper, we propose a novel Contrastive Variational Information Bottleneck framework (called CVIB) to reduce spurious correlations for ABSA. The proposed CVIB framework is composed of an original network and a self-pruned network, and these two networks are optimized simultaneously via contrastive learning. Concretely, we employ the Variational Information Bottleneck (VIB) principle to learn an informative and compressed network (self-pruned network) from the original network, which discards the superfluous patterns or spurious correlations between input features and prediction labels. Then, self-pruning contrastive learning is devised to pull together semantically similar positive pairs and push away dissimilar pairs, where the representations of the anchor learned by the original and self-pruned networks respectively are regarded as a positive pair while the representations of two different sentences within a mini-batch are treated as a negative pair. Extensive experiments on five benchmark ABSA datasets demonstrate that our CVIB method achieves better performance than the strong competitors in terms of overall prediction performance, robustness, and generalization.
☆ Model-Agnostic Meta-Learning for Natural Language Understanding Tasks in Finance
Natural language understanding(NLU) is challenging for finance due to the lack of annotated data and the specialized language in that domain. As a result, researchers have proposed to use pre-trained language model and multi-task learning to learn robust representations. However, aggressive fine-tuning often causes over-fitting and multi-task learning may favor tasks with significantly larger amounts data, etc. To address these problems, in this paper, we investigate model-agnostic meta-learning algorithm(MAML) in low-resource financial NLU tasks. Our contribution includes: 1. we explore the performance of MAML method with multiple types of tasks: GLUE datasets, SNLI, Sci-Tail and Financial PhraseBank; 2. we study the performance of MAML method with multiple single-type tasks: a real scenario stock price prediction problem with twitter text data. Our models achieve the state-of-the-art performance according to the experimental results, which demonstrate that our method can adapt fast and well to low-resource situations.
comment: 13 pages, 6 figures, 8 tables
☆ Large Language Models as Zero-Shot Human Models for Human-Robot Interaction
Human models play a crucial role in human-robot interaction (HRI), enabling robots to consider the impact of their actions on people and plan their behavior accordingly. However, crafting good human models is challenging; capturing context-dependent human behavior requires significant prior knowledge and/or large amounts of interaction data, both of which are difficult to obtain. In this work, we explore the potential of large-language models (LLMs) -- which have consumed vast amounts of human-generated text data -- to act as zero-shot human models for HRI. Our experiments on three social datasets yield promising results; the LLMs are able to achieve performance comparable to purpose-built models. That said, we also discuss current limitations, such as sensitivity to prompts and spatial/numerical reasoning mishaps. Based on our findings, we demonstrate how LLM-based human models can be integrated into a social robot's planning process and applied in HRI scenarios. Specifically, we present one case study on a simulated trust-based table-clearing task and replicate past results that relied on custom models. Next, we conduct a new robot utensil-passing experiment (n = 65) where preliminary results show that planning with a LLM-based human model can achieve gains over a basic myopic plan. In summary, our results show that LLMs offer a promising (but incomplete) approach to human modeling for HRI.
comment: 8 pages
☆ Multi-resolution Interpretation and Diagnostics Tool for Natural Language Classifiers
Developing explainability methods for Natural Language Processing (NLP) models is a challenging task, for two main reasons. First, the high dimensionality of the data (large number of tokens) results in low coverage and in turn small contributions for the top tokens, compared to the overall model performance. Second, owing to their textual nature, the input variables, after appropriate transformations, are effectively binary (presence or absence of a token in an observation), making the input-output relationship difficult to understand. Common NLP interpretation techniques do not have flexibility in resolution, because they usually operate at word-level and provide fully local (message level) or fully global (over all messages) summaries. The goal of this paper is to create more flexible model explainability summaries by segments of observation or clusters of words that are semantically related to each other. In addition, we introduce a root cause analysis method for NLP models, by analyzing representative False Positive and False Negative examples from different segments. At the end, we illustrate, using a Yelp review data set with three segments (Restaurant, Hotel, and Beauty), that exploiting group/cluster structures in words and/or messages can aid in the interpretation of decisions made by NLP models and can be utilized to assess the model's sensitivity or bias towards gender, syntax, and word meanings.
comment: 16 pages, 0 figure
☆ Guilt Detection in Text: A Step Towards Understanding Complex Emotions
We introduce a novel Natural Language Processing (NLP) task called Guilt detection, which focuses on detecting guilt in text. We identify guilt as a complex and vital emotion that has not been previously studied in NLP, and we aim to provide a more fine-grained analysis of it. To address the lack of publicly available corpora for guilt detection, we created VIC, a dataset containing 4622 texts from three existing emotion detection datasets that we binarized into guilt and no-guilt classes. We experimented with traditional machine learning methods using bag-of-words and term frequency-inverse document frequency features, achieving a 72% f1 score with the highest-performing model. Our study provides a first step towards understanding guilt in text and opens the door for future research in this area.
☆ Two-stage Pipeline for Multilingual Dialect Detection
Dialect Identification is a crucial task for localizing various Large Language Models. This paper outlines our approach to the VarDial 2023 shared task. Here we have to identify three or two dialects from three languages each which results in a 9-way classification for Track-1 and 6-way classification for Track-2 respectively. Our proposed approach consists of a two-stage system and outperforms other participants' systems and previous works in this domain. We achieve a score of 58.54% for Track-1 and 85.61% for Track-2. Our codebase is available publicly (https://github.com/ankit-vaidya19/EACL_VarDial2023).
☆ Can an Embodied Agent Find Your "Cat-shaped Mug"? LLM-Based Zero-Shot Object Navigation
We present LGX, a novel algorithm for Object Goal Navigation in a "language-driven, zero-shot manner", where an embodied agent navigates to an arbitrarily described target object in a previously unexplored environment. Our approach leverages the capabilities of Large Language Models (LLMs) for making navigational decisions by mapping the LLMs implicit knowledge about the semantic context of the environment into sequential inputs for robot motion planning. Simultaneously, we also conduct generalized target object detection using a pre-trained Vision-Language grounding model. We achieve state-of-the-art zero-shot object navigation results on RoboTHOR with a success rate (SR) improvement of over 27% over the current baseline of the OWL-ViT CLIP on Wheels (OWL CoW). Furthermore, we study the usage of LLMs for robot navigation and present an analysis of the various semantic factors affecting model output. Finally, we showcase the benefits of our approach via real-world experiments that indicate the superior performance of LGX when navigating to and detecting visually unique objects.
comment: 8 pages
☆ Spelling convention sensitivity in neural language models
We examine whether large neural language models, trained on very large collections of varied English text, learn the potentially long-distance dependency of British versus American spelling conventions, i.e., whether spelling is consistently one or the other within model-generated strings. In contrast to long-distance dependencies in non-surface underlying structure (e.g., syntax), spelling consistency is easier to measure both in LMs and the text corpora used to train them, which can provide additional insight into certain observed model behaviors. Using a set of probe words unique to either British or American English, we first establish that training corpora exhibit substantial (though not total) consistency. A large T5 language model does appear to internalize this consistency, though only with respect to observed lexical items (not nonce words with British/American spelling patterns). We further experiment with correcting for biases in the training data by fine-tuning T5 on synthetic data that has been debiased, and find that finetuned T5 remains only somewhat sensitive to spelling consistency. Further experiments show GPT2 to be similarly limited.
☆ Data Portraits: Recording Foundation Model Training Data
Foundation models are trained on increasingly immense and opaque datasets. Even while these models are now key in AI system building, it can be difficult to answer the straightforward question: has the model already encountered a given example during training? We therefore propose a widespread adoption of Data Portraits: artifacts that record training data and allow for downstream inspection. First we outline the properties of such an artifact and discuss how existing solutions can be used to increase transparency. We then propose and implement a solution based on data sketching, stressing fast and space efficient querying. Using our tool, we document a popular large language modeling corpus (the Pile) and show that our solution enables answering questions about test set leakage and model plagiarism. Our tool is lightweight and fast, costing only 3% of the dataset size in overhead. We release a demo of our tools at dataportraits.org and call on dataset and model creators to release Data Portraits as a complement to current documentation practices.
☆ Depression Detection Using Digital Traces on Social Media: A Knowledge-aware Deep Learning Approach
Depression is a common disease worldwide. It is difficult to diagnose and continues to be underdiagnosed. Because depressed patients constantly share their symptoms, major life events, and treatments on social media, researchers are turning to user-generated digital traces on social media for depression detection. Such methods have distinct advantages in combating depression because they can facilitate innovative approaches to fight depression and alleviate its social and economic burden. However, most existing studies lack effective means to incorporate established medical domain knowledge in depression detection or suffer from feature extraction difficulties that impede greater performance. Following the design science research paradigm, we propose a Deep Knowledge-aware Depression Detection (DKDD) framework to accurately detect social media users at risk of depression and explain the critical factors that contribute to such detection. Extensive empirical studies with real-world data demonstrate that, by incorporating domain knowledge, our method outperforms existing state-of-the-art methods. Our work has significant implications for IS research in knowledge-aware machine learning, digital traces utilization, and NLP research in IS. Practically, by providing early detection and explaining the critical factors, DKDD can supplement clinical depression screening and enable large-scale evaluations of a population's mental health status.
comment: Presented at INFORMS 2022 Data Science Workshop
☆ ChatGPT is on the horizon: Could a large language model be all we need for Intelligent Transportation?
ChatGPT, developed by OpenAI, is one of the largest Large Language Models (LLM) with over 175 billion parameters. ChatGPT has demonstrated the impressive capabilities of LLM, particularly in the field of natural language processing (NLP). With the emergence of the discussion and application of LLM in various research or engineering domains, it is time to envision how LLM may revolutionize the way we approach intelligent transportation systems. This paper explores the future applications of LLM in addressing key transportation problems. By leveraging LLM and a cross-modal encoder, an intelligent system can handle traffic data from various modalities and execute transportation operations through a single LLM. NLP, combined with cross-modal processing, is investigated with its potential applications in transportation. To demonstrate this potential, a smartphone-based crash report auto-generation and analysis framework is presented as a use case. Despite the potential benefits, challenges related to data privacy, data quality, and model bias must be considered. Overall, the use of LLM in intelligent transport systems holds promise for more efficient, intelligent, and sustainable transportation systems that improve the lives of people around the world.
comment: Submitted to Nature - Machine Intelligence (13 Pages, 8 Figures)
☆ wav2vec and its current potential to Automatic Speech Recognition in German for the usage in Digital History: A comparative assessment of available ASR-technologies for the use in cultural heritage contexts
In this case study we trained and published a state-of-the-art open-source model for Automatic Speech Recognition (ASR) for German to evaluate the current potential of this technology for the use in the larger context of Digital Humanities and cultural heritage indexation. Along with this paper we publish our wav2vec2 based speech to text model while we evaluate its performance on a corpus of historical recordings we assembled compared against commercial cloud-based and proprietary services. While our model achieves moderate results, we see that proprietary cloud services fare significantly better. As our results show, recognition rates over 90 percent can currently be achieved, however, these numbers drop quickly once the recordings feature limited audio quality or use of non-every day or outworn language. A big issue is the high variety of different dialects and accents in the German language. Nevertheless, this paper highlights that the currently available quality of recognition is high enough to address various use cases in the Digital Humanities. We argue that ASR will become a key technology for the documentation and analysis of audio-visual sources and identify an array of important questions that the DH community and cultural heritage stakeholders will have to address in the near future.
comment: 11 pages, 2 tables
♻ ☆ Parameter-Free Attentive Scoring for Speaker Verification
This paper presents a novel study of parameter-free attentive scoring for speaker verification. Parameter-free scoring provides the flexibility of comparing speaker representations without the need of an accompanying parametric scoring model. Inspired by the attention component in Transformer neural networks, we propose a variant of the scaled dot product attention mechanism to compare enrollment and test segment representations. In addition, this work explores the effect on performance of (i) different types of normalization, (ii) independent versus tied query/key estimation, (iii) varying the number of key-value pairs and (iv) pooling multiple enrollment utterance statistics. Experimental results for a 4 task average show that a simple parameter-free attentive scoring mechanism can improve the average EER by 10% over the best cosine similarity baseline.
♻ ☆ TalkToModel: Explaining Machine Learning Models with Interactive Natural Language Conversations
Machine Learning (ML) models are increasingly used to make critical decisions in real-world applications, yet they have become more complex, making them harder to understand. To this end, researchers have proposed several techniques to explain model predictions. However, practitioners struggle to use these explainability techniques because they often do not know which one to choose and how to interpret the results of the explanations. In this work, we address these challenges by introducing TalkToModel: an interactive dialogue system for explaining machine learning models through conversations. Specifically, TalkToModel comprises of three key components: 1) a natural language interface for engaging in conversations, making ML model explainability highly accessible, 2) a dialogue engine that adapts to any tabular model and dataset, interprets natural language, maps it to appropriate explanations, and generates text responses, and 3) an execution component that constructs the explanations. We carried out extensive quantitative and human subject evaluations of TalkToModel. Overall, we found the conversational system understands user inputs on novel datasets and models with high accuracy, demonstrating the system's capacity to generalize to new situations. In real-world evaluations with humans, 73% of healthcare workers (e.g., doctors and nurses) agreed they would use TalkToModel over baseline point-and-click systems for explainability in a disease prediction task, and 85% of ML professionals agreed TalkToModel was easier to use for computing explanations. Our findings demonstrate that TalkToModel is more effective for model explainability than existing systems, introducing a new category of explainability tools for practitioners. Code & demo released here: https://github.com/dylan-slack/TalkToModel.
comment: Pre-print; comments welcome! Reach out to dslack@uci.edu v3 update title and abstract
♻ ☆ Compositional Exemplars for In-context Learning
Large pretrained language models (LMs) have shown impressive In-Context Learning (ICL) ability, where the model learns to do an unseen task via a prompt consisting of input-output examples as the demonstration, without any parameter updates. The performance of ICL is highly dominated by the quality of the selected in-context examples. However, previous selection methods are mostly based on simple heuristics, leading to sub-optimal performance. In this work, we formulate in-context example selection as a subset selection problem. We propose CEIL (Compositional Exemplars for In-context Learning), which is instantiated by Determinantal Point Processes (DPPs) to model the interaction between the given input and in-context examples, and optimized through a carefully-designed contrastive learning objective to obtain preference from LMs. We validate CEIL on 12 classification and generation datasets from 7 distinct NLP tasks, including sentiment analysis, paraphrase detection, natural language inference, commonsense reasoning, open-domain question answering, code generation, and semantic parsing. Extensive experiments demonstrate not only the state-of-the-art performance but also the transferability and compositionality of CEIL, shedding new light on effective and efficient in-context learning. Our code is released at https://github.com/HKUNLP/icl-ceil.
♻ ☆ A Categorical Archive of ChatGPT Failures
Large language models have been demonstrated to be valuable in different fields. ChatGPT, developed by OpenAI, has been trained using massive amounts of data and simulates human conversation by comprehending context and generating appropriate responses. It has garnered significant attention due to its ability to effectively answer a broad range of human inquiries, with fluent and comprehensive answers surpassing prior public chatbots in both security and usefulness. However, a comprehensive analysis of ChatGPT's failures is lacking, which is the focus of this study. Eleven categories of failures, including reasoning, factual errors, math, coding, and bias, are presented and discussed. The risks, limitations, and societal implications of ChatGPT are also highlighted. The goal of this study is to assist researchers and developers in enhancing future language models and chatbots.
♻ ☆ Ancient Chinese Word Segmentation and Part-of-Speech Tagging Using Distant Supervision ICASSP 2023
Ancient Chinese word segmentation (WSG) and part-of-speech tagging (POS) are important to study ancient Chinese, but the amount of ancient Chinese WSG and POS tagging data is still rare. In this paper, we propose a novel augmentation method of ancient Chinese WSG and POS tagging data using distant supervision over parallel corpus. However, there are still mislabeled and unlabeled ancient Chinese words inevitably in distant supervision. To address this problem, we take advantage of the memorization effects of deep neural networks and a small amount of annotated data to get a model with much knowledge and a little noise, and then we use this model to relabel the ancient Chinese sentences in parallel corpus. Experiments show that the model trained over the relabeled data outperforms the model trained over the data generated from distant supervision and the annotated data. Our code is available at https://github.com/farlit/ACDS.
comment: Accepted by ICASSP 2023
♻ ☆ Quantifying Memorization Across Neural Language Models
Large language models (LMs) have been shown to memorize parts of their training data, and when prompted appropriately, they will emit the memorized training data verbatim. This is undesirable because memorization violates privacy (exposing user data), degrades utility (repeated easy-to-memorize text is often low quality), and hurts fairness (some texts are memorized over others). We describe three log-linear relationships that quantify the degree to which LMs emit memorized training data. Memorization significantly grows as we increase (1) the capacity of a model, (2) the number of times an example has been duplicated, and (3) the number of tokens of context used to prompt the model. Surprisingly, we find the situation becomes more complicated when generalizing these results across model families. On the whole, we find that memorization in LMs is more prevalent than previously believed and will likely get worse as models continues to scale, at least without active mitigations.
Hyena Hierarchy: Towards Larger Convolutional Language Models
Recent advances in deep learning have relied heavily on the use of large Transformers due to their ability to learn at scale. However, the core building block of Transformers, the attention operator, exhibits quadratic cost in sequence length, limiting the amount of context accessible. Existing subquadratic methods based on low-rank and sparse approximations need to be combined with dense attention layers to match Transformers, indicating a gap in capability. In this work, we propose Hyena, a subquadratic drop-in replacement for attention constructed by interleaving implicitly parametrized long convolutions and data-controlled gating. In recall and reasoning tasks on sequences of thousands to hundreds of thousands of tokens, Hyena improves accuracy by more than 50 points over operators relying on state-spaces and other implicit and explicit methods, matching attention-based models. We set a new state-of-the-art for dense-attention-free architectures on language modeling in standard datasets (WikiText103 and The Pile), reaching Transformer quality with a 20% reduction in training compute required at sequence length 2K. Hyena operators are twice as fast as highly optimized attention at sequence length 8K, and 100x faster at sequence length 64K.
comment: Additional results (PG-19, LAMBADA)
♻ ☆ DeepStruct: Pretraining of Language Models for Structure Prediction ACL 2022
We introduce a method for improving the structural understanding abilities of language models. Unlike previous approaches that finetune the models with task-specific augmentation, we pretrain language models on a collection of task-agnostic corpora to generate structures from text. Our structure pretraining enables zero-shot transfer of the learned knowledge that models have about the structure tasks. We study the performance of this approach on 28 datasets, spanning 10 structure prediction tasks including open information extraction, joint entity and relation extraction, named entity recognition, relation classification, semantic role labeling, event extraction, coreference resolution, factual probe, intent detection, and dialogue state tracking. We further enhance the pretraining with the task-specific training sets. We show that a 10B parameter language model transfers non-trivially to most tasks and obtains state-of-the-art performance on 21 of 28 datasets that we evaluate.
comment: ACL 2022
♻ ☆ Inverse scaling can become U-shaped
Scaling up language models has been empirically shown to improve performance and unlock emergent abilities. Conversely, observing worse performance as a function of scale ("inverse scaling") would indicate that scaling encourages behaviors that are misaligned with human preferences. The Inverse Scaling Prize (McKenzie et al. 2022) identified eleven such inverse scaling tasks, evaluated on models of up to 280B parameters and up to 500 zettaFLOPs of training compute. This paper takes a closer look at these inverse scaling tasks. We evaluate models of up to 540B parameters, trained on five times more compute than those evaluated in the Inverse Scaling Prize. With this increased range of model sizes and training compute, only four out of the eleven tasks remain inverse scaling. Six out of the eleven tasks exhibit what we call "U-shaped scaling" -- performance decreases up to a certain model size, and then increases again up to the largest model evaluated (the one remaining task displays positive scaling). U-shaped scaling suggests that the inverse scaling trend observed in McKenzie et al. (2022) may not continue to hold for larger models, and adds further support to the claim that sufficiently large models unlock emergent abilities.
comment: v4 reports correct results on Round 2 tasks and includes results for additional one-shot evaluation
♻ ☆ Complex QA and language models hybrid architectures, Survey
This paper provides a survey of the state of the art of hybrid language models architectures and strategies for "complex" question-answering (QA, CQA, CPS). Very large language models are good at leveraging public data on standard problems but once you want to tackle more specific complex questions or problems you may need specific architecture, knowledge, skills, tasks, methods, sensitive data, performance, human approval and versatile feedback... This survey extends findings from the robust community edited research papers BIG, BLOOM and HELM which open source, benchmark and analyze limits and challenges of large language models in terms of tasks complexity and strict evaluation on accuracy (e.g. fairness, robustness, toxicity, ...). It identifies the key elements used with Large Language Models (LLM) to solve complex questions or problems. Recent projects like ChatGPT and GALACTICA have allowed non-specialists to grasp the great potential as well as the equally strong limitations of language models in complex QA. Hybridizing these models with different components could allow to overcome these different limits and go much further. We discuss some challenges associated with complex QA, including domain adaptation, decomposition and efficient multi-step QA, long form QA, non-factoid QA, safety and multi-sensitivity data protection, multimodal search, hallucinations, QA explainability and truthfulness, time dimension. Therefore we review current solutions and promising strategies, using elements such as hybrid LLM architectures, human-in-the-loop reinforcement learning, prompting adaptation, neuro-symbolic and structured knowledge grounding, program synthesis, and others. We analyze existing solutions and provide an overview of the current research and trends in the area of complex QA.
♻ ☆ MultiViz: Towards Visualizing and Understanding Multimodal Models ICLR 2023
The promise of multimodal models for real-world applications has inspired research in visualizing and understanding their internal mechanics with the end goal of empowering stakeholders to visualize model behavior, perform model debugging, and promote trust in machine learning models. However, modern multimodal models are typically black-box neural networks, which makes it challenging to understand their internal mechanics. How can we visualize the internal modeling of multimodal interactions in these models? Our paper aims to fill this gap by proposing MultiViz, a method for analyzing the behavior of multimodal models by scaffolding the problem of interpretability into 4 stages: (1) unimodal importance: how each modality contributes towards downstream modeling and prediction, (2) cross-modal interactions: how different modalities relate with each other, (3) multimodal representations: how unimodal and cross-modal interactions are represented in decision-level features, and (4) multimodal prediction: how decision-level features are composed to make a prediction. MultiViz is designed to operate on diverse modalities, models, tasks, and research areas. Through experiments on 8 trained models across 6 real-world tasks, we show that the complementary stages in MultiViz together enable users to (1) simulate model predictions, (2) assign interpretable concepts to features, (3) perform error analysis on model misclassifications, and (4) use insights from error analysis to debug models. MultiViz is publicly available, will be regularly updated with new interpretation tools and metrics, and welcomes inputs from the community.
comment: ICLR 2023. Code available at: https://github.com/pliang279/MultiViz
Computer Vision and Pattern Recognition
☆ Detecting Human-Object Contact in Images CVPR 2023
Humans constantly contact objects to move and perform tasks. Thus, detecting human-object contact is important for building human-centered artificial intelligence. However, there exists no robust method to detect contact between the body and the scene from an image, and there exists no dataset to learn such a detector. We fill this gap with HOT ("Human-Object conTact"), a new dataset of human-object contacts for images. To build HOT, we use two data sources: (1) We use the PROX dataset of 3D human meshes moving in 3D scenes, and automatically annotate 2D image areas for contact via 3D mesh proximity and projection. (2) We use the V-COCO, HAKE and Watch-n-Patch datasets, and ask trained annotators to draw polygons for the 2D image areas where contact takes place. We also annotate the involved body part of the human body. We use our HOT dataset to train a new contact detector, which takes a single color image as input, and outputs 2D contact heatmaps as well as the body-part labels that are in contact. This is a new and challenging task that extends current foot-ground or hand-object contact detectors to the full generality of the whole body. The detector uses a part-attention branch to guide contact estimation through the context of the surrounding body parts and scene. We evaluate our detector extensively, and quantitative results show that our model outperforms baselines, and that all components contribute to better performance. Results on images from an online repository show reasonable detections and generalizability.
comment: Accepted at CVPR 2023
☆ Multimodal Prompting with Missing Modalities for Visual Recognition CVPR 2023
In this paper, we tackle two challenges in multimodal learning for visual recognition: 1) when missing-modality occurs either during training or testing in real-world situations; and 2) when the computation resources are not available to finetune on heavy transformer models. To this end, we propose to utilize prompt learning and mitigate the above two challenges together. Specifically, our modality-missing-aware prompts can be plugged into multimodal transformers to handle general missing-modality cases, while only requiring less than 1% learnable parameters compared to training the entire model. We further explore the effect of different prompt configurations and analyze the robustness to missing modality. Extensive experiments are conducted to show the effectiveness of our prompt learning framework that improves the performance under various missing-modality cases, while alleviating the requirement of heavy model re-training. Code is available.
comment: Accepted by CVPR 2023
☆ Referring Multi-Object Tracking CVPR2023
Existing referring understanding tasks tend to involve the detection of a single text-referred object. In this paper, we propose a new and general referring understanding task, termed referring multi-object tracking (RMOT). Its core idea is to employ a language expression as a semantic cue to guide the prediction of multi-object tracking. To the best of our knowledge, it is the first work to achieve an arbitrary number of referent object predictions in videos. To push forward RMOT, we construct one benchmark with scalable expressions based on KITTI, named Refer-KITTI. Specifically, it provides 18 videos with 818 expressions, and each expression in a video is annotated with an average of 10.7 objects. Further, we develop a transformer-based architecture TransRMOT to tackle the new task in an online manner, which achieves impressive detection performance and outperforms other counterparts.
comment: CVPR2023
☆ Leveraging Scene Embeddings for Gradient-Based Motion Planning in Latent Space
Motion planning framed as optimisation in structured latent spaces has recently emerged as competitive with traditional methods in terms of planning success while significantly outperforming them in terms of computational speed. However, the real-world applicability of recent work in this domain remains limited by the need to express obstacle information directly in state-space, involving simple geometric primitives. In this work we address this challenge by leveraging learned scene embeddings together with a generative model of the robot manipulator to drive the optimisation process. In addition, we introduce an approach for efficient collision checking which directly regularises the optimisation undertaken for planning. Using simulated as well as real-world experiments, we demonstrate that our approach, AMP-LS, is able to successfully plan in novel, complex scenes while outperforming traditional planning baselines in terms of computation speed by an order of magnitude. We show that the resulting system is fast enough to enable closed-loop planning in real-world dynamic scenes.
comment: Project website: https://amp-ls.github.io/
☆ Nerflets: Local Radiance Fields for Efficient Structure-Aware 3D Scene Representation from 2D Supervisio CVPR 2023
We address efficient and structure-aware 3D scene representation from images. Nerflets are our key contribution -- a set of local neural radiance fields that together represent a scene. Each nerflet maintains its own spatial position, orientation, and extent, within which it contributes to panoptic, density, and radiance reconstructions. By leveraging only photometric and inferred panoptic image supervision, we can directly and jointly optimize the parameters of a set of nerflets so as to form a decomposed representation of the scene, where each object instance is represented by a group of nerflets. During experiments with indoor and outdoor environments, we find that nerflets: (1) fit and approximate the scene more efficiently than traditional global NeRFs, (2) allow the extraction of panoptic and photometric renderings from arbitrary views, and (3) enable tasks rare for NeRFs, such as 3D panoptic segmentation and interactive editing.
comment: accepted by CVPR 2023
☆ Deep Age-Invariant Fingerprint Segmentation System
Fingerprint-based identification systems achieve higher accuracy when a slap containing multiple fingerprints of a subject is used instead of a single fingerprint. However, segmenting or auto-localizing all fingerprints in a slap image is a challenging task due to the different orientations of fingerprints, noisy backgrounds, and the smaller size of fingertip components. The presence of slap images in a real-world dataset where one or more fingerprints are rotated makes it challenging for a biometric recognition system to localize and label the fingerprints automatically. Improper fingerprint localization and finger labeling errors lead to poor matching performance. In this paper, we introduce a method to generate arbitrary angled bounding boxes using a deep learning-based algorithm that precisely localizes and labels fingerprints from both axis-aligned and over-rotated slap images. We built a fingerprint segmentation model named CRFSEG (Clarkson Rotated Fingerprint segmentation Model) by updating the previously proposed CFSEG model which was based on traditional Faster R-CNN architecture [21]. CRFSEG improves upon the Faster R-CNN algorithm with arbitrarily angled bounding boxes that allow the CRFSEG to perform better in challenging slap images. After training the CRFSEG algorithm on a new dataset containing slap images collected from both adult and children subjects, our results suggest that the CRFSEG model was invariant across different age groups and can handle over-rotated slap images successfully. In the Combined dataset containing both normal and rotated images of adult and children subjects, we achieved a matching accuracy of 97.17%, which outperformed state-of-the-art VeriFinger (94.25%) and NFSEG segmentation systems (80.58%).
comment: 20 Pages, 14 figures, Journal
☆ CleanCLIP: Mitigating Data Poisoning Attacks in Multimodal Contrastive Learning
Multimodal contrastive pretraining has been utilized to train multimodal representation models, like CLIP, on vast amounts of paired image-text data. However, previous studies have highlighted the susceptibility of such models to backdoor attacks. Specifically, when training on backdoored examples, CLIP learns spurious correlations between the embedded backdoor trigger and the target label, aligning their representations in the joint embedding space. With injecting only a few poisoned examples e.g., 75 examples in the 3M pretraining data, the model's behavior can be significantly manipulated, thus making it hard to detect or unlearn such correlations. To address this issue, we propose CleanCLIP, a finetuning framework that weakens the learned spurious associations introduced by backdoor attacks by re-aligning the representations for individual modalities independently. CleanCLIP can be employed for both unsupervised finetuning on paired image-text data and for supervised finetuning on labeled image data. We demonstrate that unsupervised finetuning with a combination of multimodal contrastive and unimodal self-supervised objectives for individual modalities can significantly reduce the impact of the backdoor attack. Additionally, supervised finetuning on task-specific labeled data of the individual modality, such as image data, removes the backdoor trigger from the CLIP vision encoder. Empirically, we show that CleanCLIP maintains model performance on benign examples while mitigating the impact of a range of backdoor attacks on multimodal contrastive learning.
comment: 20 pages, 7 figures, 8 tables
☆ MACARONS: Mapping And Coverage Anticipation with RGB Online Self-Supervision CVPR 2023
We introduce a method that simultaneously learns to explore new large environments and to reconstruct them in 3D from color images only. This is closely related to the Next Best View problem (NBV), where one has to identify where to move the camera next to improve the coverage of an unknown scene. However, most of the current NBV methods rely on depth sensors, need 3D supervision and/or do not scale to large scenes. Our method requires only a color camera and no 3D supervision. It simultaneously learns in a self-supervised fashion to predict a "volume occupancy field" from color images and, from this field, to predict the NBV. Thanks to this approach, our method performs well on new scenes as it is not biased towards any training 3D data. We demonstrate this on a recent dataset made of various 3D scenes and show it performs even better than recent methods requiring a depth sensor, which is not a realistic assumption for outdoor scenes captured with a flying drone.
comment: To appear at CVPR 2023. Project Webpage: https://imagine.enpc.fr/~guedona/MACARONS/
☆ Learning Efficient Coding of Natural Images with Maximum Manifold Capacity Representations
Self-supervised Learning (SSL) provides a strategy for constructing useful representations of images without relying on hand-assigned labels. Many such methods aim to map distinct views of the same scene or object to nearby points in the representation space, while employing some constraint to prevent representational collapse. Here we recast the problem in terms of efficient coding by adopting manifold capacity, a measure that quantifies the quality of a representation based on the number of linearly separable object manifolds it can support, as the efficiency metric to optimize. Specifically, we adapt the manifold capacity for use as an objective function in a contrastive learning framework, yielding a Maximum Manifold Capacity Representation (MMCR). We apply this method to unlabeled images, each augmented by a set of basic transformations, and find that it learns meaningful features using the standard linear evaluation protocol. Specifically, we find that MMCRs support performance on object recognition comparable to or surpassing that of recently developed SSL frameworks, while providing more robustness to adversarial attacks. Empirical analyses reveal differences between MMCRs and representations learned by other SSL frameworks, and suggest a mechanism by which manifold compression gives rise to class separability.
☆ Exploring Deep Models for Practical Gait Recognition
Gait recognition is a rapidly advancing vision technique for person identification from a distance. Prior studies predominantly employed relatively small and shallow neural networks to extract subtle gait features, achieving impressive successes in indoor settings. Nevertheless, experiments revealed that these existing methods mostly produce unsatisfactory results when applied to newly released in-the-wild gait datasets. This paper presents a unified perspective to explore how to construct deep models for state-of-the-art outdoor gait recognition, including the classical CNN-based and emerging Transformer-based architectures. Consequently, we emphasize the importance of suitable network capacity, explicit temporal modeling, and deep transformer structure for discriminative gait representation learning. Our proposed CNN-based DeepGaitV2 series and Transformer-based SwinGait series exhibit significant performance gains in outdoor scenarios, \textit{e.g.}, about +30\% rank-1 accuracy compared with many state-of-the-art methods on the challenging GREW dataset. This work is expected to further boost the research and application of gait recognition. Code will be available at https://github.com/ShiqiYu/OpenGait.
☆ Visual Place Recognition: A Tutorial
Localization is an essential capability for mobile robots. A rapidly growing field of research in this area is Visual Place Recognition (VPR), which is the ability to recognize previously seen places in the world based solely on images. This present work is the first tutorial paper on visual place recognition. It unifies the terminology of VPR and complements prior research in two important directions: 1) It provides a systematic introduction for newcomers to the field, covering topics such as the formulation of the VPR problem, a general-purpose algorithmic pipeline, an evaluation methodology for VPR approaches, and the major challenges for VPR and how they may be addressed. 2) As a contribution for researchers acquainted with the VPR problem, it examines the intricacies of different VPR problem types regarding input, data processing, and output. The tutorial also discusses the subtleties behind the evaluation of VPR algorithms, e.g., the evaluation of a VPR system that has to find all matching database images per query, as opposed to just a single match. Practical code examples in Python illustrate to prospective practitioners and researchers how VPR is implemented and evaluated.
comment: Under review
☆ Evaluating the Fairness of Deep Learning Uncertainty Estimates in Medical Image Analysis
Although deep learning (DL) models have shown great success in many medical image analysis tasks, deployment of the resulting models into real clinical contexts requires: (1) that they exhibit robustness and fairness across different sub-populations, and (2) that the confidence in DL model predictions be accurately expressed in the form of uncertainties. Unfortunately, recent studies have indeed shown significant biases in DL models across demographic subgroups (e.g., race, sex, age) in the context of medical image analysis, indicating a lack of fairness in the models. Although several methods have been proposed in the ML literature to mitigate a lack of fairness in DL models, they focus entirely on the absolute performance between groups without considering their effect on uncertainty estimation. In this work, we present the first exploration of the effect of popular fairness models on overcoming biases across subgroups in medical image analysis in terms of bottom-line performance, and their effects on uncertainty quantification. We perform extensive experiments on three different clinically relevant tasks: (i) skin lesion classification, (ii) brain tumour segmentation, and (iii) Alzheimer's disease clinical score regression. Our results indicate that popular ML methods, such as data-balancing and distributionally robust optimization, succeed in mitigating fairness issues in terms of the model performances for some of the tasks. However, this can come at the cost of poor uncertainty estimates associated with the model predictions. This tradeoff must be mitigated if fairness models are to be adopted in medical image analysis.
comment: Paper accepted at MIDL 2023
☆ StyO: Stylize Your Face in Only One-Shot
This paper focuses on face stylization with a single artistic target. Existing works for this task often fail to retain the source content while achieving geometry variation. Here, we present a novel StyO model, ie. Stylize the face in only One-shot, to solve the above problem. In particular, StyO exploits a disentanglement and recombination strategy. It first disentangles the content and style of source and target images into identifiers, which are then recombined in a cross manner to derive the stylized face image. In this way, StyO decomposes complex images into independent and specific attributes, and simplifies one-shot face stylization as the combination of different attributes from input images, thus producing results better matching face geometry of target image and content of source one. StyO is implemented with latent diffusion models (LDM) and composed of two key modules: 1) Identifier Disentanglement Learner (IDL) for disentanglement phase. It represents identifiers as contrastive text prompts, ie. positive and negative descriptions. And it introduces a novel triple reconstruction loss to fine-tune the pre-trained LDM for encoding style and content into corresponding identifiers; 2) Fine-grained Content Controller (FCC) for the recombination phase. It recombines disentangled identifiers from IDL to form an augmented text prompt for generating stylized faces. In addition, FCC also constrains the cross-attention maps of latent and text features to preserve source face details in results. The extensive evaluation shows that StyO produces high-quality images on numerous paintings of various styles and outperforms the current state-of-the-art. Code will be released upon acceptance.
☆ Combination of Single and Multi-frame Image Super-resolution: An Analytical Perspective
Super-resolution is the process of obtaining a high-resolution image from one or more low-resolution images. Single image super-resolution (SISR) and multi-frame super-resolution (MFSR) methods have been evolved almost independently for years. A neglected study in this field is the theoretical analysis of finding the optimum combination of SISR and MFSR. To fill this gap, we propose a novel theoretical analysis based on the iterative shrinkage and thresholding algorithm. We implement and compare several approaches for combining SISR and MFSR, and simulation results support the finding of our theoretical analysis, both quantitatively and qualitatively.
☆ Continuous Sign Language Recognition with Correlation Network CVPR2023
Human body trajectories are a salient cue to identify actions in the video. Such body trajectories are mainly conveyed by hands and face across consecutive frames in sign language. However, current methods in continuous sign language recognition (CSLR) usually process frames independently, thus failing to capture cross-frame trajectories to effectively identify a sign. To handle this limitation, we propose correlation network (CorrNet) to explicitly capture and leverage body trajectories across frames to identify signs. In specific, a correlation module is first proposed to dynamically compute correlation maps between the current frame and adjacent frames to identify trajectories of all spatial patches. An identification module is then presented to dynamically emphasize the body trajectories within these correlation maps. As a result, the generated features are able to gain an overview of local temporal movements to identify a sign. Thanks to its special attention on body trajectories, CorrNet achieves new state-of-the-art accuracy on four large-scale datasets, i.e., PHOENIX14, PHOENIX14-T, CSL-Daily, and CSL. A comprehensive comparison with previous spatial-temporal reasoning methods verifies the effectiveness of CorrNet. Visualizations demonstrate the effects of CorrNet on emphasizing human body trajectories across adjacent frames.
comment: CVPR2023, code: https://github.com/hulianyuyy/CorrNet. arXiv admin note: text overlap with arXiv:2211.17081
☆ A System for Generalized 3D Multi-Object Search ICRA
Searching for objects is a fundamental skill for robots. As such, we expect object search to eventually become an off-the-shelf capability for robots, similar to e.g., object detection and SLAM. In contrast, however, no system for 3D object search exists that generalizes across real robots and environments. In this paper, building upon a recent theoretical framework that exploited the octree structure for representing belief in 3D, we present GenMOS (Generalized Multi-Object Search), the first general-purpose system for multi-object search (MOS) in a 3D region that is robot-independent and environment-agnostic. GenMOS takes as input point cloud observations of the local region, object detection results, and localization of the robot's view pose, and outputs a 6D viewpoint to move to through online planning. In particular, GenMOS uses point cloud observations in three ways: (1) to simulate occlusion; (2) to inform occupancy and initialize octree belief; and (3) to sample a belief-dependent graph of view positions that avoid obstacles. We evaluate our system both in simulation and on two real robot platforms. Our system enables, for example, a Boston Dynamics Spot robot to find a toy cat hidden underneath a couch in under one minute. We further integrate 3D local search with 2D global search to handle larger areas, demonstrating the resulting system in a 25m$^2$ lobby area.
comment: 8 pages, 9 figures, 1 table. IEEE Conference on Robotics and Automation (ICRA) 2023
☆ Neighborhood Contrastive Transformer for Change Captioning
Change captioning is to describe the semantic change between a pair of similar images in natural language. It is more challenging than general image captioning, because it requires capturing fine-grained change information while being immune to irrelevant viewpoint changes, and solving syntax ambiguity in change descriptions. In this paper, we propose a neighborhood contrastive transformer to improve the model's perceiving ability for various changes under different scenes and cognition ability for complex syntax structure. Concretely, we first design a neighboring feature aggregating to integrate neighboring context into each feature, which helps quickly locate the inconspicuous changes under the guidance of conspicuous referents. Then, we devise a common feature distilling to compare two images at neighborhood level and extract common properties from each image, so as to learn effective contrastive information between them. Finally, we introduce the explicit dependencies between words to calibrate the transformer decoder, which helps better understand complex syntax structure during training. Extensive experimental results demonstrate that the proposed method achieves the state-of-the-art performance on three public datasets with different change scenarios. The code is available at https://github.com/tuyunbin/NCT.
comment: Accepted by IEEE TMM
☆ Faster Learning of Temporal Action Proposal via Sparse Multilevel Boundary Generator
Temporal action localization in videos presents significant challenges in the field of computer vision. While the boundary-sensitive method has been widely adopted, its limitations include incomplete use of intermediate and global information, as well as an inefficient proposal feature generator. To address these challenges, we propose a novel framework, Sparse Multilevel Boundary Generator (SMBG), which enhances the boundary-sensitive method with boundary classification and action completeness regression. SMBG features a multi-level boundary module that enables faster processing by gathering boundary information at different lengths. Additionally, we introduce a sparse extraction confidence head that distinguishes information inside and outside the action, further optimizing the proposal feature generator. To improve the synergy between multiple branches and balance positive and negative samples, we propose a global guidance loss. Our method is evaluated on two popular benchmarks, ActivityNet-1.3 and THUMOS14, and is shown to achieve state-of-the-art performance, with a better inference speed (2.47xBSN++, 2.12xDBG). These results demonstrate that SMBG provides a more efficient and simple solution for generating temporal action proposals. Our proposed framework has the potential to advance the field of computer vision and enhance the accuracy and speed of temporal action localization in video analysis.The code and models are made available at \url{https://github.com/zhouyang-001/SMBG-for-temporal-action-proposal}.
comment: 18 pages, 5 figures
☆ Video Question Answering Using CLIP-Guided Visual-Text Attention
Cross-modal learning of video and text plays a key role in Video Question Answering (VideoQA). In this paper, we propose a visual-text attention mechanism to utilize the Contrastive Language-Image Pre-training (CLIP) trained on lots of general domain language-image pairs to guide the cross-modal learning for VideoQA. Specifically, we first extract video features using a TimeSformer and text features using a BERT from the target application domain, and utilize CLIP to extract a pair of visual-text features from the general-knowledge domain through the domain-specific learning. We then propose a Cross-domain Learning to extract the attention information between visual and linguistic features across the target domain and general domain. The set of CLIP-guided visual-text features are integrated to predict the answer. The proposed method is evaluated on MSVD-QA and MSRVTT-QA datasets, and outperforms state-of-the-art methods.
☆ ST-KeyS: Self-Supervised Transformer for Keyword Spotting in Historical Handwritten Documents
Keyword spotting (KWS) in historical documents is an important tool for the initial exploration of digitized collections. Nowadays, the most efficient KWS methods are relying on machine learning techniques that require a large amount of annotated training data. However, in the case of historical manuscripts, there is a lack of annotated corpus for training. To handle the data scarcity issue, we investigate the merits of the self-supervised learning to extract useful representations of the input data without relying on human annotations and then using these representations in the downstream task. We propose ST-KeyS, a masked auto-encoder model based on vision transformers where the pretraining stage is based on the mask-and-predict paradigm, without the need of labeled data. In the fine-tuning stage, the pre-trained encoder is integrated into a siamese neural network model that is fine-tuned to improve feature embedding from the input images. We further improve the image representation using pyramidal histogram of characters (PHOC) embedding to create and exploit an intermediate representation of images based on text attributes. In an exhaustive experimental evaluation on three widely used benchmark datasets (Botany, Alvermann Konzilsprotokolle and George Washington), the proposed approach outperforms state-of-the-art methods trained on the same datasets.
☆ CRIN: Rotation-Invariant Point Cloud Analysis and Rotation Estimation via Centrifugal Reference Frame AAAI 2023
Various recent methods attempt to implement rotation-invariant 3D deep learning by replacing the input coordinates of points with relative distances and angles. Due to the incompleteness of these low-level features, they have to undertake the expense of losing global information. In this paper, we propose the CRIN, namely Centrifugal Rotation-Invariant Network. CRIN directly takes the coordinates of points as input and transforms local points into rotation-invariant representations via centrifugal reference frames. Aided by centrifugal reference frames, each point corresponds to a discrete rotation so that the information of rotations can be implicitly stored in point features. Unfortunately, discrete points are far from describing the whole rotation space. We further introduce a continuous distribution for 3D rotations based on points. Furthermore, we propose an attention-based down-sampling strategy to sample points invariant to rotations. A relation module is adopted at last for reinforcing the long-range dependencies between sampled points and predicts the anchor point for unsupervised rotation estimation. Extensive experiments show that our method achieves rotation invariance, accurately estimates the object rotation, and obtains state-of-the-art results on rotation-augmented classification and part segmentation. Ablation studies validate the effectiveness of the network design.
comment: AAAI 2023
☆ MOISST: Multi-modal Optimization of Implicit Scene for SpatioTemporal calibration
With the recent advances in autonomous driving and the decreasing cost of LiDARs, the use of multi-modal sensor systems is on the rise. However, in order to make use of the information provided by a variety of complimentary sensors, it is necessary to accurately calibrate them. We take advantage of recent advances in computer graphics and implicit volumetric scene representation to tackle the problem of multi-sensor spatial and temporal calibration. Thanks to a new formulation of the implicit model optimization, we are able to jointly optimize calibration parameters along with scene representation based on radiometric and geometric measurements. Our method enables accurate and robust calibration from data captured in uncontrolled and unstructured urban environments, making our solution more scalable than existing calibration solutions. We demonstrate the accuracy and robustness of our method in urban scenes typically encountered in autonomous driving scenarios.
☆ Masked Images Are Counterfactual Samples for Robust Fine-tuning CVPR 2023
Deep learning models are challenged by the distribution shift between the training data and test data. Recently, the large models pre-trained on diverse data demonstrate unprecedented robustness to various distribution shifts. However, fine-tuning on these models can lead to a trade-off between in-distribution (ID) performance and out-of-distribution (OOD) robustness. Existing methods for tackling this trade-off do not explicitly address the OOD robustness problem. In this paper, based on causal analysis on the aforementioned problems, we propose a novel fine-tuning method, which use masked images as counterfactual samples that help improving the robustness of the fine-tuning model. Specifically, we mask either the semantics-related or semantics-unrelated patches of the images based on class activation map to break the spurious correlation, and refill the masked patches with patches from other images. The resulting counterfactual samples are used in feature-based distillation with the pre-trained model. Extensive experiments verify that regularizing the fine-tuning with the proposed masked images can achieve a better trade-off between ID and OOD, surpassing previous methods on the OOD performance. Our code will be publicly available.
comment: Accepted by CVPR 2023
☆ MABNet: Master Assistant Buddy Network with Hybrid Learning for Image Retrieval ICASSP
Image retrieval has garnered growing interest in recent times. The current approaches are either supervised or self-supervised. These methods do not exploit the benefits of hybrid learning using both supervision and self-supervision. We present a novel Master Assistant Buddy Network (MABNet) for image retrieval which incorporates both learning mechanisms. MABNet consists of master and assistant blocks, both learning independently through supervision and collectively via self-supervision. The master guides the assistant by providing its knowledge base as a reference for self-supervision and the assistant reports its knowledge back to the master by weight transfer. We perform extensive experiments on public datasets with and without post-processing.
comment: Accepted at International Conference on Acoustics, Speech, and Signal Processing (ICASSP) 2023
☆ Automatic detection of aerial survey ground control points based on Yolov5-OBB
The use of ground control points (GCPs) for georeferencing is the most common strategy in unmanned aerial vehicle (UAV) photogrammetry, but at the same time their collection represents the most time-consuming and expensive part of UAV campaigns. Recently, deep learning has been rapidly developed in the field of small object detection. In this letter, to automatically extract coordinates information of ground control points (GCPs) by detecting GCP-markers in UAV images, we propose a solution that uses a deep learning-based architecture, YOLOv5-OBB, combined with a confidence threshold filtering algorithm and an optimal ranking algorithm. We applied our proposed method to a dataset collected by DJI Phantom 4 Pro drone and obtained good detection performance with the mean Average Precision (AP) of 0.832 and the highest AP of 0.982 for the cross-type GCP-markers. The proposed method can be a promising tool for future implementation of the end-to-end aerial triangulation process.
comment: 6 pages, 4 figures
☆ EvCenterNet: Uncertainty Estimation for Object Detection using Evidential Learning
Uncertainty estimation is crucial in safety-critical settings such as automated driving as it provides valuable information for several downstream tasks including high-level decision-making and path planning. In this work, we propose EvCenterNet, a novel uncertainty-aware 2D object detection framework utilizing evidential learning to directly estimate both classification and regression uncertainties. To employ evidential learning for object detection, we devise a combination of evidential and focal loss functions for the sparse heatmap inputs. We introduce class-balanced weighting for regression and heatmap prediction to tackle the class imbalance encountered by evidential learning. Moreover, we propose a learning scheme to actively utilize the predicted heatmap uncertainties to improve the detection performance by focusing on the most uncertain points. We train our model on the KITTI dataset and evaluate it on challenging out-of-distribution datasets including BDD100K and nuImages. Our experiments demonstrate that our approach improves the precision and minimizes the execution time loss in relation to the base model.
☆ DeCap: Decoding CLIP Latents for Zero-Shot Captioning via Text-Only Training ICLR 2023
Large-scale pre-trained multi-modal models (e.g., CLIP) demonstrate strong zero-shot transfer capability in many discriminative tasks. Their adaptation to zero-shot image-conditioned text generation tasks has drawn increasing interest. Prior arts approach to zero-shot captioning by either utilizing the existing large language models (e.g., GPT-2) or pre-training the encoder-decoder network in an end-to-end manner. In this work, we propose a simple framework, named DeCap, for zero-shot captioning. We introduce a lightweight visual-aware language decoder. This decoder is both data-efficient and computation-efficient: 1) it only requires the text data for training, easing the burden on the collection of paired data. 2) it does not require end-to-end training. When trained with text-only data, the decoder takes the text embedding extracted from the off-the-shelf CLIP encoder as a prefix embedding. The challenge is that the decoder is trained on the text corpus but at the inference stage, it needs to generate captions based on visual inputs. The modality gap issue is widely observed in multi-modal contrastive models that prevents us from directly taking the visual embedding as the prefix embedding. We propose a training-free mechanism to reduce the modality gap. We project the visual embedding into the CLIP text embedding space, while the projected embedding retains the information of the visual input. Taking the projected embedding as the prefix embedding, the decoder generates high-quality descriptions that match the visual input. The experiments show that DeCap outperforms other zero-shot captioning methods and unpaired captioning methods on the typical image captioning benchmarks, i.e., MSCOCO and NoCaps.
comment: Accepted by ICLR 2023. Code is available at https://github.com/dhg-wei/DeCap
☆ RQAT-INR: Improved Implicit Neural Image Compression
Deep variational autoencoders for image and video compression have gained significant attraction in the recent years, due to their potential to offer competitive or better compression rates compared to the decades long traditional codecs such as AVC, HEVC or VVC. However, because of complexity and energy consumption, these approaches are still far away from practical usage in industry. More recently, implicit neural representation (INR) based codecs have emerged, and have lower complexity and energy usage to classical approaches at decoding. However, their performances are not in par at the moment with state-of-the-art methods. In this research, we first show that INR based image codec has a lower complexity than VAE based approaches, then we propose several improvements for INR-based image codec and outperformed baseline model by a large margin.
comment: Accepted as oral at Data compression conference 2023
☆ Guiding Energy-based Models via Contrastive Latent Variables ICLR 2023
An energy-based model (EBM) is a popular generative framework that offers both explicit density and architectural flexibility, but training them is difficult since it is often unstable and time-consuming. In recent years, various training techniques have been developed, e.g., better divergence measures or stabilization in MCMC sampling, but there often exists a large gap between EBMs and other generative frameworks like GANs in terms of generation quality. In this paper, we propose a novel and effective framework for improving EBMs via contrastive representation learning (CRL). To be specific, we consider representations learned by contrastive methods as the true underlying latent variable. This contrastive latent variable could guide EBMs to understand the data structure better, so it can improve and accelerate EBM training significantly. To enable the joint training of EBM and CRL, we also design a new class of latent-variable EBMs for learning the joint density of data and the contrastive latent variable. Our experimental results demonstrate that our scheme achieves lower FID scores, compared to prior-art EBM methods (e.g., additionally using variational autoencoders or diffusion techniques), even with significantly faster and more memory-efficient training. We also show conditional and compositional generation abilities of our latent-variable EBMs as their additional benefits, even without explicit conditional training. The code is available at https://github.com/hankook/CLEL.
comment: Accepted to ICLR 2023 (Spotlight). The code is available at https://github.com/hankook/CLEL
☆ Efficient Large-scale Scene Representation with a Hybrid of High-resolution Grid and Plane Features
Existing neural radiance fields (NeRF) methods for large-scale scene modeling require days of training using multiple GPUs, hindering their applications in scenarios with limited computing resources. Despite fast optimization NeRF variants have been proposed based on the explicit dense or hash grid features, their effectivenesses are mainly demonstrated in object-scale scene representation. In this paper, we point out that the low feature resolution in explicit representation is the bottleneck for large-scale unbounded scene representation. To address this problem, we introduce a new and efficient hybrid feature representation for NeRF that fuses the 3D hash-grids and high-resolution 2D dense plane features. Compared with the dense-grid representation, the resolution of a dense 2D plane can be scaled up more efficiently. Based on this hybrid representation, we propose a fast optimization NeRF variant, called GP-NeRF, that achieves better rendering results while maintaining a compact model size. Extensive experiments on multiple large-scale unbounded scene datasets show that our model can converge in 1.5 hours using a single GPU while achieving results comparable to or even better than the existing method that requires about one day's training with 8 GPUs.
☆ Pseudo-label Correction and Learning For Semi-Supervised Object Detection
Pseudo-Labeling has emerged as a simple yet effective technique for semi-supervised object detection (SSOD). However, the inevitable noise problem in pseudo-labels significantly degrades the performance of SSOD methods. Recent advances effectively alleviate the classification noise in SSOD, while the localization noise which is a non-negligible part of SSOD is not well-addressed. In this paper, we analyse the localization noise from the generation and learning phases, and propose two strategies, namely pseudo-label correction and noise-unaware learning. For pseudo-label correction, we introduce a multi-round refining method and a multi-vote weighting method. The former iteratively refines the pseudo boxes to improve the stability of predictions, while the latter smoothly self-corrects pseudo boxes by weighing the scores of surrounding jittered boxes. For noise-unaware learning, we introduce a loss weight function that is negatively correlated with the Intersection over Union (IoU) in the regression task, which pulls the predicted boxes closer to the object and improves localization accuracy. Our proposed method, Pseudo-label Correction and Learning (PCL), is extensively evaluated on the MS COCO and PASCAL VOC benchmarks. On MS COCO, PCL outperforms the supervised baseline by 12.16, 12.11, and 9.57 mAP and the recent SOTA (SoftTeacher) by 3.90, 2.54, and 2.43 mAP under 1\%, 5\%, and 10\% labeling ratios, respectively. On PASCAL VOC, PCL improves the supervised baseline by 5.64 mAP and the recent SOTA (Unbiased Teacherv2) by 1.04 mAP on AP$^{50}$.
☆ HiCLIP: Contrastive Language-Image Pretraining with Hierarchy-aware Attention ICLR 2023
The success of large-scale contrastive vision-language pretraining (CLIP) has benefited both visual recognition and multimodal content understanding. The concise design brings CLIP the advantage in inference efficiency against other vision-language models with heavier cross-attention fusion layers, making it a popular choice for a wide spectrum of downstream tasks. However, CLIP does not explicitly capture the hierarchical nature of high-level and fine-grained semantics conveyed in images and texts, which is arguably critical to vision-language understanding and reasoning. To this end, we equip both the visual and language branches in CLIP with hierarchy-aware attentions, namely Hierarchy-aware CLIP (HiCLIP), to progressively discover semantic hierarchies layer-by-layer from both images and texts in an unsupervised manner. As a result, such hierarchical aggregation significantly improves the cross-modal alignment. To demonstrate the advantages of HiCLIP, we conduct qualitative analysis on its unsupervised hierarchy induction during inference, as well as extensive quantitative experiments on both visual recognition and vision-language downstream tasks.
comment: Accepted at ICLR 2023
☆ Fighting noise and imbalance in Action Unit detection problems
Action Unit (AU) detection aims at automatically caracterizing facial expressions with the muscular activations they involve. Its main interest is to provide a low-level face representation that can be used to assist higher level affective computing tasks learning. Yet, it is a challenging task. Indeed, the available databases display limited face variability and are imbalanced toward neutral expressions. Furthermore, as AU involve subtle face movements they are difficult to annotate so that some of the few provided datapoints may be mislabeled. In this work, we aim at exploiting label smoothing ability to mitigate noisy examples impact by reducing confidence [1]. However, applying label smoothing as it is may aggravate imbalance-based pre-existing under-confidence issue and degrade performance. To circumvent this issue, we propose Robin Hood Label Smoothing (RHLS). RHLS principle is to restrain label smoothing confidence reduction to the majority class. In that extent, it alleviates both the imbalance-based over-confidence issue and the negative impact of noisy majority class examples. From an experimental standpoint, we show that RHLS provides a free performance improvement in AU detection. In particular, by applying it on top of a modern multi-task baseline we get promising results on BP4D and outperform state-of-the-art methods on DISFA.
☆ Learning multi-scale local conditional probability models of images
Deep neural networks can learn powerful prior probability models for images, as evidenced by the high-quality generations obtained with recent score-based diffusion methods. But the means by which these networks capture complex global statistical structure, apparently without suffering from the curse of dimensionality, remain a mystery. To study this, we incorporate diffusion methods into a multi-scale decomposition, reducing dimensionality by assuming a stationary local Markov model for wavelet coefficients conditioned on coarser-scale coefficients. We instantiate this model using convolutional neural networks (CNNs) with local receptive fields, which enforce both the stationarity and Markov properties. Global structures are captured using a CNN with receptive fields covering the entire (but small) low-pass image. We test this model on a dataset of face images, which are highly non-stationary and contain large-scale geometric structures. Remarkably, denoising, super-resolution, and image synthesis results all demonstrate that these structures can be captured with significantly smaller conditioning neighborhoods than required by a Markov model implemented in the pixel domain. Our results show that score estimation for large complex images can be reduced to low-dimensional Markov conditional models across scales, alleviating the curse of dimensionality.
comment: 16 pages, 8 figures
☆ CLIP-guided Prototype Modulating for Few-shot Action Recognition
Learning from large-scale contrastive language-image pre-training like CLIP has shown remarkable success in a wide range of downstream tasks recently, but it is still under-explored on the challenging few-shot action recognition (FSAR) task. In this work, we aim to transfer the powerful multimodal knowledge of CLIP to alleviate the inaccurate prototype estimation issue due to data scarcity, which is a critical problem in low-shot regimes. To this end, we present a CLIP-guided prototype modulating framework called CLIP-FSAR, which consists of two key components: a video-text contrastive objective and a prototype modulation. Specifically, the former bridges the task discrepancy between CLIP and the few-shot video task by contrasting videos and corresponding class text descriptions. The latter leverages the transferable textual concepts from CLIP to adaptively refine visual prototypes with a temporal Transformer. By this means, CLIP-FSAR can take full advantage of the rich semantic priors in CLIP to obtain reliable prototypes and achieve accurate few-shot classification. Extensive experiments on five commonly used benchmarks demonstrate the effectiveness of our proposed method, and CLIP-FSAR significantly outperforms existing state-of-the-art methods under various settings. The source code and models will be publicly available at https://github.com/alibaba-mmai-research/CLIP-FSAR.
comment: This work has been submitted to the Springer for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessible
☆ System for 3D Acquisition and 3D Reconstruction using Structured Light for Sewer Line Inspection
The assessment of sewer pipe systems is a highly important, but at the same time cumbersome and error-prone task. We introduce an innovative system based on single-shot structured light modules that facilitates the detection and classification of spatial defects like jutting intrusions, spallings, or misaligned joints. This system creates highly accurate 3D measurements with sub-millimeter resolution of pipe surfaces and fuses them into a holistic 3D model. The benefit of such a holistic 3D model is twofold: on the one hand, it facilitates the accurate manual sewer pipe assessment, on the other, it simplifies the detection of defects in downstream automatic systems as it endows the input with highly accurate depth information. In this work, we provide an extensive overview of the system and give valuable insights into our design choices.
comment: 10 pages, published at VISAPP 2023, Lisbon, Portugal
☆ Histogram-based Deep Learning for Automotive Radar
There are various automotive applications that rely on correctly interpreting point cloud data recorded with radar sensors. We present a deep learning approach for histogram-based processing of such point clouds. Compared to existing methods, the design of our approach is extremely simple: it boils down to computing a point cloud histogram and passing it through a multi-layer perceptron. Our approach matches and surpasses state-of-the-art approaches on the task of automotive radar object type classification. It is also robust to noise that often corrupts radar measurements, and can deal with missing features of single radar reflections. Finally, the design of our approach makes it more interpretable than existing methods, allowing insightful analysis of its decisions.
☆ Rethinking Confidence Calibration for Failure Prediction ECCV 2022
Reliable confidence estimation for the predictions is important in many safety-critical applications. However, modern deep neural networks are often overconfident for their incorrect predictions. Recently, many calibration methods have been proposed to alleviate the overconfidence problem. With calibrated confidence, a primary and practical purpose is to detect misclassification errors by filtering out low-confidence predictions (known as failure prediction). In this paper, we find a general, widely-existed but actually-neglected phenomenon that most confidence calibration methods are useless or harmful for failure prediction. We investigate this problem and reveal that popular confidence calibration methods often lead to worse confidence separation between correct and incorrect samples, making it more difficult to decide whether to trust a prediction or not. Finally, inspired by the natural connection between flat minima and confidence separation, we propose a simple hypothesis: flat minima is beneficial for failure prediction. We verify this hypothesis via extensive experiments and further boost the performance by combining two different flat minima techniques. Our code is available at https://github.com/Impression2805/FMFP
comment: Accepted to ECCV 2022. Code is available at https://github.com/Impression2805/FMFP
☆ DwinFormer: Dual Window Transformers for End-to-End Monocular Depth Estimation
Depth estimation from a single image is of paramount importance in the realm of computer vision, with a multitude of applications. Conventional methods suffer from the trade-off between consistency and fine-grained details due to the local-receptive field limiting their practicality. This lack of long-range dependency inherently comes from the convolutional neural network part of the architecture. In this paper, a dual window transformer-based network, namely DwinFormer, is proposed, which utilizes both local and global features for end-to-end monocular depth estimation. The DwinFormer consists of dual window self-attention and cross-attention transformers, Dwin-SAT and Dwin-CAT, respectively. The Dwin-SAT seamlessly extracts intricate, locally aware features while concurrently capturing global context. It harnesses the power of local and global window attention to adeptly capture both short-range and long-range dependencies, obviating the need for complex and computationally expensive operations, such as attention masking or window shifting. Moreover, Dwin-SAT introduces inductive biases which provide desirable properties, such as translational equvariance and less dependence on large-scale data. Furthermore, conventional decoding methods often rely on skip connections which may result in semantic discrepancies and a lack of global context when fusing encoder and decoder features. In contrast, the Dwin-CAT employs both local and global window cross-attention to seamlessly fuse encoder and decoder features with both fine-grained local and contextually aware global information, effectively amending semantic gap. Empirical evidence obtained through extensive experimentation on the NYU-Depth-V2 and KITTI datasets demonstrates the superiority of the proposed method, consistently outperforming existing approaches across both indoor and outdoor environments.
☆ Automated Peripancreatic Vessel Segmentation and Labeling Based on Iterative Trunk Growth and Weakly Supervised Mechanism
Peripancreatic vessel segmentation and anatomical labeling play extremely important roles to assist the early diagnosis, surgery planning and prognosis for patients with pancreatic tumors. However, most current techniques cannot achieve satisfactory segmentation performance for peripancreatic veins and usually make predictions with poor integrity and connectivity. Besides, unsupervised labeling algorithms cannot deal with complex anatomical variation while fully supervised methods require a large number of voxel-wise annotations for training, which is very labor-intensive and time-consuming. To address these problems, we propose our Automated Peripancreatic vEssel Segmentation and lAbeling (APESA) framework, to not only highly improve the segmentation performance for peripancreatic veins, but also efficiently identify the peripancreatic artery branches. There are two core modules in our proposed APESA framework: iterative trunk growth module (ITGM) for vein segmentation and weakly supervised labeling mechanism (WSLM) for artery branch identification. Our proposed ITGM is composed of a series of trunk growth modules, each of which chooses the most reliable trunk of a basic vessel prediction by the largest connected constraint, and seeks for the possible growth branches by branch proposal network. Our designed iterative process guides the raw trunk to be more complete and fully connected. Our proposed WSLM consists of an unsupervised rule-based preprocessing for generating pseudo branch annotations, and an anatomical labeling network to learn the branch distribution voxel by voxel. We achieve Dice of 94.01% for vein segmentation on our collected dataset, which boosts the accuracy by nearly 10% compared with the state-of-the-art methods. Additionally, we also achieve Dice of 97.01% on segmentation and competitive performance on anatomical labeling for peripancreatic arteries.
☆ Models See Hallucinations: Evaluating the Factuality in Video Captioning
Video captioning aims to describe events in a video with natural language. In recent years, many works have focused on improving captioning models' performance. However, like other text generation tasks, it risks introducing factual errors not supported by the input video. These factual errors can seriously affect the quality of the generated text, sometimes making it completely unusable. Although factual consistency has received much research attention in text-to-text tasks (e.g., summarization), it is less studied in the context of vision-based text generation. In this work, we conduct a detailed human evaluation of the factuality in video captioning and collect two annotated factuality datasets. We find that 57.0% of the model-generated sentences have factual errors, indicating it is a severe problem in this field. However, existing evaluation metrics are mainly based on n-gram matching and show little correlation with human factuality annotation. We further propose a weakly-supervised, model-based factuality metric FactVC, which outperforms previous metrics on factuality evaluation of video captioning. The datasets and metrics will be released to promote future research for video captioning.
comment: 17 pages
☆ Butterfly: Multiple Reference Frames Feature Propagation Mechanism for Neural Video Compression
Using more reference frames can significantly improve the compression efficiency in neural video compression. However, in low-latency scenarios, most existing neural video compression frameworks usually use the previous one frame as reference. Or a few frameworks which use the previous multiple frames as reference only adopt a simple multi-reference frames propagation mechanism. In this paper, we present a more reasonable multi-reference frames propagation mechanism for neural video compression, called butterfly multi-reference frame propagation mechanism (Butterfly), which allows a more effective feature fusion of multi-reference frames. By this, we can generate more accurate temporal context conditional prior for Contextual Coding Module. Besides, when the number of decoded frames does not meet the required number of reference frames, we duplicate the nearest reference frame to achieve the requirement, which is better than duplicating the furthest one. Experiment results show that our method can significantly outperform the previous state-of-the-art (SOTA), and our neural codec can achieve -7.6% bitrate save on HEVC Class D dataset when compares with our base single-reference frame model with the same compression configuration.
comment: Accepted by DCC 2023
☆ Centroid Distance Distillation for Effective Rehearsal in Continual Learning
Rehearsal, retraining on a stored small data subset of old tasks, has been proven effective in solving catastrophic forgetting in continual learning. However, due to the sampled data may have a large bias towards the original dataset, retraining them is susceptible to driving continual domain drift of old tasks in feature space, resulting in forgetting. In this paper, we focus on tackling the continual domain drift problem with centroid distance distillation. First, we propose a centroid caching mechanism for sampling data points based on constructed centroids to reduce the sample bias in rehearsal. Then, we present a centroid distance distillation that only stores the centroid distance to reduce the continual domain drift. The experiments on four continual learning datasets show the superiority of the proposed method, and the continual domain drift can be reduced.
☆ CTG-Net: An Efficient Cascaded Framework Driven by Terminal Guidance Mechanism for Dilated Pancreatic Duct Segmentation
Pancreatic duct dilation indicates a high risk of various pancreatic diseases. Segmentation of dilated pancreatic ducts on computed tomography (CT) images shows the potential to assist the early diagnosis, surgical planning and prognosis. Because of the ducts' tiny sizes, slender tubular structures and the surrounding distractions, most current researches on pancreatic duct segmentation achieve low accuracy and always have segmentation errors on the terminal parts of the ducts. To address these problems, we propose a terminal guidance mechanism called cascaded terminal guidance network (CTG-Net). Firstly, a terminal attention mechanism is established on the skeletons extracted from the coarse predictions. Then, to get fine terminal segmentation, a subnetwork is designed for jointly learning the local intensity from the original images, feature cues from coarse predictions and global anatomy information from the pancreas distance transform maps. Finally, a terminal distraction attention module which explicitly learns the distribution of the terminal distraction is proposed to reduce the false positive and false negative predictions. We also propose a new metric called tDice to measure the terminal segmentation accuracy for targets with tubular structures and two segmentation metrics for distractions. We collect our dilated pancreatic duct segmentation dataset with 150 CT scans from patients with 5 types of pancreatic tumors. Experimental results on our dataset show that our proposed approach boosts dilated pancreatic duct segmentation accuracy by nearly 20% compared with the existing results, and achieves more than 9% improvement for the terminal segmentation accuracy compared with the state-of-the-art methods.
☆ Adaptive Texture Filtering for Single-Domain Generalized Segmentation AAAI 2023
Domain generalization in semantic segmentation aims to alleviate the performance degradation on unseen domains through learning domain-invariant features. Existing methods diversify images in the source domain by adding complex or even abnormal textures to reduce the sensitivity to domain specific features. However, these approaches depend heavily on the richness of the texture bank, and training them can be time-consuming. In contrast to importing textures arbitrarily or augmenting styles randomly, we focus on the single source domain itself to achieve generalization. In this paper, we present a novel adaptive texture filtering mechanism to suppress the influence of texture without using augmentation, thus eliminating the interference of domain-specific features. Further, we design a hierarchical guidance generalization network equipped with structure-guided enhancement modules, which purpose is to learn the domain-invariant generalized knowledge. Extensive experiments together with ablation studies on widely-used datasets are conducted to verify the effectiveness of the proposed model, and reveal its superiority over other state-of-the-art alternatives.
comment: Accepted by AAAI 2023
☆ UniHCP: A Unified Model for Human-Centric Perceptions CVPR 2023
Human-centric perceptions (e.g., pose estimation, human parsing, pedestrian detection, person re-identification, etc.) play a key role in industrial applications of visual models. While specific human-centric tasks have their own relevant semantic aspect to focus on, they also share the same underlying semantic structure of the human body. However, few works have attempted to exploit such homogeneity and design a general-propose model for human-centric tasks. In this work, we revisit a broad range of human-centric tasks and unify them in a minimalist manner. We propose UniHCP, a Unified Model for Human-Centric Perceptions, which unifies a wide range of human-centric tasks in a simplified end-to-end manner with the plain vision transformer architecture. With large-scale joint training on 33 human-centric datasets, UniHCP can outperform strong baselines on several in-domain and downstream tasks by direct evaluation. When adapted to a specific task, UniHCP achieves new SOTAs on a wide range of human-centric tasks, e.g., 69.8 mIoU on CIHP for human parsing, 86.18 mA on PA-100K for attribute prediction, 90.3 mAP on Market1501 for ReID, and 85.8 JI on CrowdHuman for pedestrian detection, performing better than specialized models tailored for each task.
comment: Accepted for publication at the IEEE/CVF Conference on Computer Vision and Pattern Recognition 2023 (CVPR 2023)
☆ Scapegoat Generation for Privacy Protection from Deepfake
To protect privacy and prevent malicious use of deepfake, current studies propose methods that interfere with the generation process, such as detection and destruction approaches. However, these methods suffer from sub-optimal generalization performance to unseen models and add undesirable noise to the original image. To address these problems, we propose a new problem formulation for deepfake prevention: generating a ``scapegoat image'' by modifying the style of the original input in a way that is recognizable as an avatar by the user, but impossible to reconstruct the real face. Even in the case of malicious deepfake, the privacy of the users is still protected. To achieve this, we introduce an optimization-based editing method that utilizes GAN inversion to discourage deepfake models from generating similar scapegoats. We validate the effectiveness of our proposed method through quantitative and user studies.
comment: 5 pages, 5 figures
☆ SurfNN: Joint Reconstruction of Multiple Cortical Surfaces from Magnetic Resonance Images
To achieve fast, robust, and accurate reconstruction of the human cortical surfaces from 3D magnetic resonance images (MRIs), we develop a novel deep learning-based framework, referred to as SurfNN, to reconstruct simultaneously both inner (between white matter and gray matter) and outer (pial) surfaces from MRIs. Different from existing deep learning-based cortical surface reconstruction methods that either reconstruct the cortical surfaces separately or neglect the interdependence between the inner and outer surfaces, SurfNN reconstructs both the inner and outer cortical surfaces jointly by training a single network to predict a midthickness surface that lies at the center of the inner and outer cortical surfaces. The input of SurfNN consists of a 3D MRI and an initialization of the midthickness surface that is represented both implicitly as a 3D distance map and explicitly as a triangular mesh with spherical topology, and its output includes both the inner and outer cortical surfaces, as well as the midthickness surface. The method has been evaluated on a large-scale MRI dataset and demonstrated competitive cortical surface reconstruction performance.
comment: ISBI 2023
☆ MotionVideoGAN: A Novel Video Generator Based on the Motion Space Learned from Image Pairs
Video generation has achieved rapid progress benefiting from high-quality renderings provided by powerful image generators. We regard the video synthesis task as generating a sequence of images sharing the same contents but varying in motions. However, most previous video synthesis frameworks based on pre-trained image generators treat content and motion generation separately, leading to unrealistic generated videos. Therefore, we design a novel framework to build the motion space, aiming to achieve content consistency and fast convergence for video generation. We present MotionVideoGAN, a novel video generator synthesizing videos based on the motion space learned by pre-trained image pair generators. Firstly, we propose an image pair generator named MotionStyleGAN to generate image pairs sharing the same contents and producing various motions. Then we manage to acquire motion codes to edit one image in the generated image pairs and keep the other unchanged. The motion codes help us edit images within the motion space since the edited image shares the same contents with the other unchanged one in image pairs. Finally, we introduce a latent code generator to produce latent code sequences using motion codes for video generation. Our approach achieves state-of-the-art performance on the most complex video dataset ever used for unconditional video generation evaluation, UCF101.
comment: Accepted by IEEE Transactions on Multimedia as a regular paper
☆ Improving Transformer-based Image Matching by Cascaded Capturing Spatially Informative Keypoints
Learning robust local image feature matching is a fundamental low-level vision task, which has been widely explored in the past few years. Recently, detector-free local feature matchers based on transformers have shown promising results, which largely outperform pure Convolutional Neural Network (CNN) based ones. But correlations produced by transformer-based methods are spatially limited to the center of source views' coarse patches, because of the costly attention learning. In this work, we rethink this issue and find that such matching formulation degrades pose estimation, especially for low-resolution images. So we propose a transformer-based cascade matching model -- Cascade feature Matching TRansformer (CasMTR), to efficiently learn dense feature correlations, which allows us to choose more reliable matching pairs for the relative pose estimation. Instead of re-training a new detector, we use a simple yet effective Non-Maximum Suppression (NMS) post-process to filter keypoints through the confidence map, and largely improve the matching precision. CasMTR achieves state-of-the-art performance in indoor and outdoor pose estimation as well as visual localization. Moreover, thorough ablations show the efficacy of the proposed components and techniques.
☆ KBNet: Kernel Basis Network for Image Restoration
How to aggregate spatial information plays an essential role in learning-based image restoration. Most existing CNN-based networks adopt static convolutional kernels to encode spatial information, which cannot aggregate spatial information adaptively. Recent transformer-based architectures achieve adaptive spatial aggregation. But they lack desirable inductive biases of convolutions and require heavy computational costs. In this paper, we propose a kernel basis attention (KBA) module, which introduces learnable kernel bases to model representative image patterns for spatial information aggregation. Different kernel bases are trained to model different local structures. At each spatial location, they are linearly and adaptively fused by predicted pixel-wise coefficients to obtain aggregation weights. Based on the KBA module, we further design a multi-axis feature fusion (MFF) block to encode and fuse channel-wise, spatial-invariant, and pixel-adaptive features for image restoration. Our model, named kernel basis network (KBNet), achieves state-of-the-art performances on more than ten benchmarks over image denoising, deraining, and deblurring tasks while requiring less computational cost than previous SOTA methods.
☆ Spatiotemporal Capsule Neural Network for Vehicle Trajectory Prediction
Through advancement of the Vehicle-to-Everything (V2X) network, road safety, energy consumption, and traffic efficiency can be significantly improved. An accurate vehicle trajectory prediction benefits communication traffic management and network resource allocation for the real-time application of the V2X network. Recurrent neural networks and their variants have been reported in recent research to predict vehicle mobility. However, the spatial attribute of vehicle movement behavior has been overlooked, resulting in incomplete information utilization. To bridge this gap, we put forward for the first time a hierarchical trajectory prediction structure using the capsule neural network (CapsNet) with three sequential components. First, the geographic information is transformed into a grid map presentation, describing vehicle mobility distribution spatially and temporally. Second, CapsNet serves as the core model to embed local temporal and global spatial correlation through hierarchical capsules. Finally, extensive experiments conducted on actual taxi mobility data collected in Porto city (Portugal) and Singapore show that the proposed method outperforms the state-of-the-art methods.
comment: IEEE TVT has accepted this paper
☆ A Review of Deep Learning-Powered Mesh Reconstruction Methods
With the recent advances in hardware and rendering techniques, 3D models have emerged everywhere in our life. Yet creating 3D shapes is arduous and requires significant professional knowledge. Meanwhile, Deep learning has enabled high-quality 3D shape reconstruction from various sources, making it a viable approach to acquiring 3D shapes with minimal effort. Importantly, to be used in common 3D applications, the reconstructed shapes need to be represented as polygonal meshes, which is a challenge for neural networks due to the irregularity of mesh tessellations. In this survey, we provide a comprehensive review of mesh reconstruction methods that are powered by machine learning. We first describe various representations for 3D shapes in the deep learning context. Then we review the development of 3D mesh reconstruction methods from voxels, point clouds, single images, and multi-view images. Finally, we identify several challenges in this field and propose potential future directions.
☆ Enhancing Border Security and Countering Terrorism Through Computer Vision: a Field of Artificial Intelligence
Border security had been a persistent problem in international border especially when it get to the issue of preventing illegal movement of weapons, contraband, drugs, and combating issue of illegal or undocumented immigrant while at the same time ensuring that lawful trade, economic prosperity coupled with national sovereignty across the border is maintained. In this research work, we used open source computer vision (Open CV) and adaboost algorithm to develop a model which can detect a moving object a far off, classify it, automatically snap full image and face of the individual separately, and then run a background check on them against worldwide databases while making a prediction about an individual being a potential threat, intending immigrant, potential terrorists or extremist and then raise sound alarm. Our model can be deployed on any camera device and be mounted at any international border. There are two stages involved, we first developed a model based on open CV computer vision algorithm, with the ability to detect human movement from afar, it will automatically snap both the face and the full image of the person separately, and the second stage is the automatic triggering of background check against the moving object. This ensures it check the moving object against several databases worldwide and is able to determine the admissibility of the person afar off. If the individual is inadmissible, it will automatically alert the border officials with the image of the person and other details, and if the bypass the border officials, the system is able to detect and alert the authority with his images and other details. All these operations will be done afar off by the AI powered camera before the individual reach the border
comment: 10 pages, 8 figures, Conference publication
☆ Dual Feedback Attention Framework via Boundary-Aware Auxiliary and Progressive Semantic Optimization for Salient Object Detection in Optical Remote Sensing Imagery
Salient object detection in optical remote sensing image (ORSI-SOD) has gradually attracted attention thanks to the development of deep learning (DL) and salient object detection in natural scene image (NSI-SOD). However, NSI and ORSI are different in many aspects, such as large coverage, complex background, and large differences in target types and scales. Therefore, a new dedicated method is needed for ORSI-SOD. In addition, existing methods do not pay sufficient attention to the boundary of the object, and the completeness of the final saliency map still needs improvement. To address these issues, we propose a novel method called Dual Feedback Attention Framework via Boundary-Aware Auxiliary and Progressive Semantic Optimization (DFA-BASO). First, Boundary Protection Calibration (BPC) module is proposed to reduce the loss of edge position information during forward propagation and suppress noise in low-level features. Second, a Dual Feature Feedback Complementary (DFFC) module is proposed based on BPC module. It aggregates boundary-semantic dual features and provides effective feedback to coordinate features across different layers. Finally, a Strong Semantic Feedback Refinement (SSFR) module is proposed to obtain more complete saliency maps. This module further refines feature representation and eliminates feature differences through a unique feedback mechanism. Extensive experiments on two public datasets show that DFA-BASO outperforms 15 state-of-the-art methods. Furthermore, this paper strongly demonstrates the true contribution of DFA-BASO to ORSI-SOD by in-depth analysis of the visualization figure. All codes can be found at https://github.com/YUHsss/DFA-BASO.
☆ EvHandPose: Event-based 3D Hand Pose Estimation with Sparse Supervision
Event camera shows great potential in 3D hand pose estimation, especially addressing the challenges of fast motion and high dynamic range in a low-power way. However, due to the asynchronous differential imaging mechanism, it is challenging to design event representation to encode hand motion information especially when the hands are not moving (causing motion ambiguity), and it is infeasible to fully annotate the temporally dense event stream. In this paper, we propose EvHandPose with novel hand flow representations in Event-to-Pose module for accurate hand pose estimation and alleviating the motion ambiguity issue. To solve the problem under sparse annotation, we design contrast maximization and edge constraints in Pose-to-IWE (Image with Warped Events) module and formulate EvHandPose in a self-supervision framework. We further build EvRealHands, the first large-scale real-world event-based hand pose dataset on several challenging scenes to bridge the domain gap due to relying on synthetic data and facilitate future research. Experiments on EvRealHands demonstrate that EvHandPose outperforms previous event-based method under all evaluation scenes with 15 $\sim$ 20 mm lower MPJPE and achieves accurate and stable hand pose estimation in fast motion and strong light scenes compared with RGB-based methods. Furthermore, EvHandPose demonstrates 3D hand pose estimation at 120 fps or higher.
☆ Weakly Supervised Realtime Dynamic Background Subtraction
Background subtraction is a fundamental task in computer vision with numerous real-world applications, ranging from object tracking to video surveillance. Dynamic backgrounds poses a significant challenge here. Supervised deep learning-based techniques are currently considered state-of-the-art for this task. However, these methods require pixel-wise ground-truth labels, which can be time-consuming and expensive. In this work, we propose a weakly supervised framework that can perform background subtraction without requiring per-pixel ground-truth labels. Our framework is trained on a moving object-free sequence of images and comprises two networks. The first network is an autoencoder that generates background images and prepares dynamic background images for training the second network. The dynamic background images are obtained by thresholding the background-subtracted images. The second network is a U-Net that uses the same object-free video for training and the dynamic background images as pixel-wise ground-truth labels. During the test phase, the input images are processed by the autoencoder and U-Net, which generate background and dynamic background images, respectively. The dynamic background image helps remove dynamic motion from the background-subtracted image, enabling us to obtain a foreground image that is free of dynamic artifacts. To demonstrate the effectiveness of our method, we conducted experiments on selected categories of the CDnet 2014 dataset and the I2R dataset. Our method outperformed all top-ranked unsupervised methods. We also achieved better results than one of the two existing weakly supervised methods, and our performance was similar to the other. Our proposed method is online, real-time, efficient, and requires minimal frame-level annotation, making it suitable for a wide range of real-world applications.
comment: 10 pages, 3 figures
☆ Traffic Scene Parsing through the TSP6K Dataset
Traffic scene parsing is one of the most important tasks to achieve intelligent cities. So far, little effort has been spent on constructing datasets specifically for the task of traffic scene parsing. To fill this gap, here we introduce the TSP6K dataset, containing 6,000 urban traffic images and spanning hundreds of street scenes under various weather conditions. In contrast to most previous traffic scene datasets collected from a driving platform, the images in our dataset are from the shooting platform high-hanging on the street. Such traffic images can capture more crowded street scenes with several times more traffic participants than the driving scenes. Each image in the TSP6K dataset is provided with high-quality pixel-level and instance-level annotations. We perform a detailed analysis for the dataset and comprehensively evaluate the state-of-the-art scene parsing methods. Considering the vast difference in instance sizes, we propose a detail refining decoder, which recovers the details of different semantic regions in traffic scenes. Experiments have shown its effectiveness in parsing high-hanging traffic scenes. Code and dataset will be made publicly available.
comment: 11 pages, 7 figures
☆ Robust Autoencoders for Collective Corruption Removal ICASSP2023
Robust PCA is a standard tool for learning a linear subspace in the presence of sparse corruption or rare outliers. What about robustly learning manifolds that are more realistic models for natural data, such as images? There have been several recent attempts to generalize robust PCA to manifold settings. In this paper, we propose $\ell_1$- and scaling-invariant $\ell_1/\ell_2$-robust autoencoders based on a surprisingly compact formulation built on the intuition that deep autoencoders perform manifold learning. We demonstrate on several standard image datasets that the proposed formulation significantly outperforms all previous methods in collectively removing sparse corruption, without clean images for training. Moreover, we also show that the learned manifold structures can be generalized to unseen data samples effectively.
comment: This paper has been accepted to ICASSP2023
☆ Visual Analytics of Neuron Vulnerability to Adversarial Attacks on Convolutional Neural Networks
Adversarial attacks on a convolutional neural network (CNN) -- injecting human-imperceptible perturbations into an input image -- could fool a high-performance CNN into making incorrect predictions. The success of adversarial attacks raises serious concerns about the robustness of CNNs, and prevents them from being used in safety-critical applications, such as medical diagnosis and autonomous driving. Our work introduces a visual analytics approach to understanding adversarial attacks by answering two questions: (1) which neurons are more vulnerable to attacks and (2) which image features do these vulnerable neurons capture during the prediction? For the first question, we introduce multiple perturbation-based measures to break down the attacking magnitude into individual CNN neurons and rank the neurons by their vulnerability levels. For the second, we identify image features (e.g., cat ears) that highly stimulate a user-selected neuron to augment and validate the neuron's responsibility. Furthermore, we support an interactive exploration of a large number of neurons by aiding with hierarchical clustering based on the neurons' roles in the prediction. To this end, a visual analytics system is designed to incorporate visual reasoning for interpreting adversarial attacks. We validate the effectiveness of our system through multiple case studies as well as feedback from domain experts.
comment: Accepted by the Special Issue on Human-Centered Explainable AI, ACM Transactions on Interactive Intelligent Systems
☆ Hyperspectral Compressive Wavefront Sensing
Presented is a novel way to combine snapshot compressive imaging and lateral shearing interferometry in order to capture the spatio-spectral phase of an ultrashort laser pulse in a single shot. A deep unrolling algorithm is utilised for the snapshot compressive imaging reconstruction due to its parameter efficiency and superior speed relative to other methods, potentially allowing for online reconstruction. The algorithm's regularisation term is represented using neural network with 3D convolutional layers, to exploit the spatio-spectral correlations that exist in laser wavefronts. Compressed sensing is not typically applied to modulated signals, but we demonstrate its success here. Furthermore, we train a neural network to predict the wavefronts from a lateral shearing interferogram in terms of Zernike polynomials, which again increases the speed of our technique without sacrificing fidelity. This method is supported with simulation-based results. While applied to the example of lateral shearing interferometry, the methods presented here are generally applicable to a wide range of signals, including Shack-Hartmann-type sensors. The results may be of interest beyond the context of laser wavefront characterization, including within quantitative phase imaging.
☆ Memory Maps for Video Object Detection and Tracking on UAVs
This paper introduces a novel approach to video object detection detection and tracking on Unmanned Aerial Vehicles (UAVs). By incorporating metadata, the proposed approach creates a memory map of object locations in actual world coordinates, providing a more robust and interpretable representation of object locations in both, image space and the real world. We use this representation to boost confidences, resulting in improved performance for several temporal computer vision tasks, such as video object detection, short and long-term single and multi-object tracking, and video anomaly detection. These findings confirm the benefits of metadata in enhancing the capabilities of UAVs in the field of temporal computer vision and pave the way for further advancements in this area.
☆ Refining 3D Human Texture Estimation from a Single Image
Estimating 3D human texture from a single image is essential in graphics and vision. It requires learning a mapping function from input images of humans with diverse poses into the parametric (UV) space and reasonably hallucinating invisible parts. To achieve a high-quality 3D human texture estimation, we propose a framework that adaptively samples the input by a deformable convolution where offsets are learned via a deep neural network. Additionally, we describe a novel cycle consistency loss that improves view generalization. We further propose to train our framework with an uncertainty-based pixel-level image reconstruction loss, which enhances color fidelity. We compare our method against the state-of-the-art approaches and show significant qualitative and quantitative improvements.
☆ Towards Composable Distributions of Latent Space Augmentations
We propose a composable framework for latent space image augmentation that allows for easy combination of multiple augmentations. Image augmentation has been shown to be an effective technique for improving the performance of a wide variety of image classification and generation tasks. Our framework is based on the Variational Autoencoder architecture and uses a novel approach for augmentation via linear transformation within the latent space itself. We explore losses and augmentation latent geometry to enforce the transformations to be composable and involuntary, thus allowing the transformations to be readily combined or inverted. Finally, we show these properties are better performing with certain pairs of augmentations, but we can transfer the latent space to other sets of augmentations to modify performance, effectively constraining the VAE's bottleneck to preserve the variance of specific augmentations and features of the image which we care about. We demonstrate the effectiveness of our approach with initial results on the MNIST dataset against both a standard VAE and a Conditional VAE. This latent augmentation method allows for much greater control and geometric interpretability of the latent space, making it a valuable tool for researchers and practitioners in the field.
comment: Accepted at 2023 Information Theory and Applications Workshop (Feb, San Diego)
☆ Learning Differential Invariants of Planar Curves
We propose a learning paradigm for the numerical approximation of differential invariants of planar curves. Deep neural-networks' (DNNs) universal approximation properties are utilized to estimate geometric measures. The proposed framework is shown to be a preferable alternative to axiomatic constructions. Specifically, we show that DNNs can learn to overcome instabilities and sampling artifacts and produce consistent signatures for curves subject to a given group of transformations in the plane. We compare the proposed schemes to alternative state-of-the-art axiomatic constructions of differential invariants. We evaluate our models qualitatively and quantitatively and propose a benchmark dataset to evaluate approximation models of differential invariants of planar curves.
comment: SSVM 2023. arXiv admin note: substantial text overlap with arXiv:2202.05922
☆ Polar Prediction of Natural Videos
Observer motion and continuous deformations of objects and surfaces imbue natural videos with distinct temporal structures, enabling partial prediction of future frames from past ones. Conventional methods first estimate local motion, or optic flow, and then use it to predict future frames by warping or copying content. Here, we explore a more direct methodology, in which each frame is mapped into a learned representation space where the structure of temporal evolution is more readily accessible. Motivated by the geometry of the Fourier shift theorem and its group-theoretic generalization, we formulate a simple architecture that represents video frames in learned local polar coordinates. Specifically, we construct networks in which pairs of convolutional channel coefficients are treated as complex-valued, and are optimized to evolve with slowly varying amplitudes and linearly advancing phases. We train these models on next-frame prediction in natural videos, and compare their performance with that of conventional methods using optic flow as well as predictive neural networks. We find that the polar predictor achieves better performance while remaining interpretable and fast, thereby demonstrating the potential of a flow-free video processing methodology that is trained end-to-end to predict natural video content.
☆ Neural Style Transfer for Vector Graphics
Neural style transfer draws researchers' attention, but the interest focuses on bitmap images. Various models have been developed for bitmap image generation both online and offline with arbitrary and pre-trained styles. However, the style transfer between vector images has not almost been considered. Our research shows that applying standard content and style losses insignificantly changes the vector image drawing style because the structure of vector primitives differs a lot from pixels. To handle this problem, we introduce new loss functions. We also develop a new method based on differentiable rasterization that uses these loss functions and can change the color and shape parameters of the content image corresponding to the drawing of the style image. Qualitative experiments demonstrate the effectiveness of the proposed VectorNST method compared with the state-of-the-art neural style transfer approaches for bitmap images and the only existing approach for stylizing vector images, DiffVG. Although the proposed model does not achieve the quality and smoothness of style transfer between bitmap images, we consider our work an important early step in this area. VectorNST code and demo service are available at https://github.com/IzhanVarsky/VectorNST.
♻ ☆ All are Worth Words: A ViT Backbone for Diffusion Models
Vision transformers (ViT) have shown promise in various vision tasks while the U-Net based on a convolutional neural network (CNN) remains dominant in diffusion models. We design a simple and general ViT-based architecture (named U-ViT) for image generation with diffusion models. U-ViT is characterized by treating all inputs including the time, condition and noisy image patches as tokens and employing long skip connections between shallow and deep layers. We evaluate U-ViT in unconditional and class-conditional image generation, as well as text-to-image generation tasks, where U-ViT is comparable if not superior to a CNN-based U-Net of a similar size. In particular, latent diffusion models with U-ViT achieve record-breaking FID scores of 2.29 in class-conditional image generation on ImageNet 256x256, and 5.48 in text-to-image generation on MS-COCO, among methods without accessing large external datasets during the training of generative models. Our results suggest that, for diffusion-based image modeling, the long skip connection is crucial while the down-sampling and up-sampling operators in CNN-based U-Net are not always necessary. We believe that U-ViT can provide insights for future research on backbones in diffusion models and benefit generative modeling on large scale cross-modality datasets.
♻ ☆ Neural Congealing: Aligning Images to a Joint Semantic Atlas
We present Neural Congealing -- a zero-shot self-supervised framework for detecting and jointly aligning semantically-common content across a given set of images. Our approach harnesses the power of pre-trained DINO-ViT features to learn: (i) a joint semantic atlas -- a 2D grid that captures the mode of DINO-ViT features in the input set, and (ii) dense mappings from the unified atlas to each of the input images. We derive a new robust self-supervised framework that optimizes the atlas representation and mappings per image set, requiring only a few real-world images as input without any additional input information (e.g., segmentation masks). Notably, we design our losses and training paradigm to account only for the shared content under severe variations in appearance, pose, background clutter or other distracting objects. We demonstrate results on a plethora of challenging image sets including sets of mixed domains (e.g., aligning images depicting sculpture and artwork of cats), sets depicting related yet different object categories (e.g., dogs and tigers), or domains for which large-scale training data is scarce (e.g., coffee mugs). We thoroughly evaluate our method and show that our test-time optimization approach performs favorably compared to a state-of-the-art method that requires extensive training on large-scale datasets.
comment: Project page: https://neural-congealing.github.io/
♻ ☆ RRWaveNet: A Compact End-to-End Multi-Scale Residual CNN for Robust PPG Respiratory Rate Estimation
Respiratory rate (RR) is an important biomarker as RR changes can reflect severe medical events such as heart disease, lung disease, and sleep disorders. Unfortunately, standard manual RR counting is prone to human error and cannot be performed continuously. This study proposes a method for continuously estimating RR, RRWaveNet. The method is a compact end-to-end deep learning model which does not require feature engineering and can use low-cost raw photoplethysmography (PPG) as input signal. RRWaveNet was tested subject-independently and compared to baseline in four datasets (BIDMC, CapnoBase, WESAD, and SensAI) and using three window sizes (16, 32, and 64 seconds). RRWaveNet outperformed current state-of-the-art methods with mean absolute errors at optimal window size of 1.66 \pm 1.01, 1.59 \pm 1.08, 1.92 \pm 0.96 and 1.23 \pm 0.61 breaths per minute for each dataset. In remote monitoring settings, such as in the WESAD and SensAI datasets, we apply transfer learning to improve the performance using two other ICU datasets as pretraining datasets, reducing the MAE by up to 21$\%$. This shows that this model allows accurate and practical estimation of RR on affordable and wearable devices. Our study also shows feasibility of remote RR monitoring in the context of telemedicine and at home.
comment: 11 pages, 8 figures
♻ ☆ Word-As-Image for Semantic Typography
A word-as-image is a semantic typography technique where a word illustration presents a visualization of the meaning of the word, while also preserving its readability. We present a method to create word-as-image illustrations automatically. This task is highly challenging as it requires semantic understanding of the word and a creative idea of where and how to depict these semantics in a visually pleasing and legible manner. We rely on the remarkable ability of recent large pretrained language-vision models to distill textual concepts visually. We target simple, concise, black-and-white designs that convey the semantics clearly. We deliberately do not change the color or texture of the letters and do not use embellishments. Our method optimizes the outline of each letter to convey the desired concept, guided by a pretrained Stable Diffusion model. We incorporate additional loss terms to ensure the legibility of the text and the preservation of the style of the font. We show high quality and engaging results on numerous examples and compare to alternative techniques.
♻ ☆ PaintNet: Unstructured Multi-Path Learning from 3D Point Clouds for Robotic Spray Painting
Popular industrial robotic problems such as spray painting and welding require (i) conditioning on free-shape 3D objects and (ii) planning of multiple trajectories to solve the task. Yet, existing solutions make strong assumptions on the form of input surfaces and the nature of output paths, resulting in limited approaches unable to cope with real-data variability. By leveraging on recent advances in 3D deep learning, we introduce a novel framework capable of dealing with arbitrary 3D surfaces, and handling a variable number of unordered output paths (i.e. unstructured). Our approach focuses on predicting smaller path segments, which can be later concatenated to reconstruct long-horizon paths. We extensively validate the proposed method in the context of robotic spray painting by releasing PaintNet, the first public dataset of expert demonstrations on free-shape 3D objects collected in a real industrial scenario. A thorough experimental analysis demonstrates the capabilities of our model to promptly predict smooth output paths that cover up to 95% of the surface of previously unseen object instances. Furthermore, we show how models learned from PaintNet capture relevant features which serve as a reliable starting point to improve data and time efficiency when dealing with new object categories.
comment: Project website at https://gabrieletiboni.github.io/paintnet
♻ ☆ Bootstrapping Semi-supervised Medical Image Segmentation with Anatomical-aware Contrastive Distillation
Contrastive learning has shown great promise over annotation scarcity problems in the context of medical image segmentation. Existing approaches typically assume a balanced class distribution for both labeled and unlabeled medical images. However, medical image data in reality is commonly imbalanced (i.e., multi-class label imbalance), which naturally yields blurry contours and usually incorrectly labels rare objects. Moreover, it remains unclear whether all negative samples are equally negative. In this work, we present ACTION, an Anatomical-aware ConTrastive dIstillatiON framework, for semi-supervised medical image segmentation. Specifically, we first develop an iterative contrastive distillation algorithm by softly labeling the negatives rather than binary supervision between positive and negative pairs. We also capture more semantically similar features from the randomly chosen negative set compared to the positives to enforce the diversity of the sampled data. Second, we raise a more important question: Can we really handle imbalanced samples to yield better performance? Hence, the key innovation in ACTION is to learn global semantic relationship across the entire dataset and local anatomical features among the neighbouring pixels with minimal additional memory footprint. During the training, we introduce anatomical contrast by actively sampling a sparse set of hard negative pixels, which can generate smoother segmentation boundaries and more accurate predictions. Extensive experiments across two benchmark datasets and different unlabeled settings show that ACTION significantly outperforms the current state-of-the-art semi-supervised methods.
♻ ☆ Joint cortical registration of geometry and function using semi-supervised learning
Brain surface-based image registration, an important component of brain image analysis, establishes spatial correspondence between cortical surfaces. Existing iterative and learning-based approaches focus on accurate registration of folding patterns of the cerebral cortex, and assume that geometry predicts function and thus functional areas will also be well aligned. However, structure/functional variability of anatomically corresponding areas across subjects has been widely reported. In this work, we introduce a learning-based cortical registration framework, JOSA, which jointly aligns folding patterns and functional maps while simultaneously learning an optimal atlas. We demonstrate that JOSA can substantially improve registration performance in both anatomical and functional domains over existing methods. By employing a semi-supervised training strategy, the proposed framework obviates the need for functional data during inference, enabling its use in broad neuroscientific domains where functional data may not be observed.
comment: * co-senior authors with equal contribution
♻ ☆ Enhancing Fine-Grained 3D Object Recognition using Hybrid Multi-Modal Vision Transformer-CNN Models
Robots operating in human-centered environments, such as retail stores, restaurants, and households, are often required to distinguish between similar objects in different contexts with a high degree of accuracy. However, fine-grained object recognition remains a challenge in robotics due to the high intra-category and low inter-category dissimilarities. In addition, the limited number of fine-grained 3D datasets poses a significant problem in addressing this issue effectively. In this paper, we propose a hybrid multi-modal Vision Transformer (ViT) and Convolutional Neural Networks (CNN) approach to improve the performance of fine-grained visual classification (FGVC). To address the shortage of FGVC 3D datasets, we generated two synthetic datasets. The first dataset consists of 20 categories related to restaurants with a total of 100 instances, while the second dataset contains 120 shoe instances. Our approach was evaluated on both datasets, and the results indicate that it outperforms both CNN-only and ViT-only baselines, achieving a recognition accuracy of 94.50 % and 93.51 % on the restaurant and shoe datasets, respectively. Additionally, we have made our FGVC RGB-D datasets available to the research community to enable further experimentation and advancement. Furthermore, we successfully integrated our proposed method with a robot framework and demonstrated its potential as a fine-grained perception tool in both simulated and real-world robotic scenarios.
♻ ☆ Evaluation of Interpretability Methods and Perturbation Artifacts in Deep Neural Networks
Despite excellent performance of deep neural networks (DNNs) in image classification, detection, and prediction, characterizing how DNNs make a given decision remains an open problem, resulting in a number of interpretability methods. Post-hoc interpretability methods primarily aim to quantify the importance of input features with respect to the class probabilities. However, due to the lack of ground truth and the existence of interpretability methods with diverse operating characteristics, evaluating these methods is a crucial challenge. A popular approach to evaluate interpretability methods is to perturb input features deemed important for a given prediction and observe the decrease in accuracy. However, perturbation itself may introduce artifacts, since perturbed images may be out-of-distribution (OOD). In this paper, we have conducted computational experiments to estimate the contribution of perturbation artifacts and developed a method to estimate the fidelity of interpretability methods. We demonstrate that, while perturbation artifacts indeed exist, we can minimize and characterize their impact on fidelity estimation by utilizing model accuracy curves from perturbing input features according to the Most Import First (MIF) and Least Import First (LIF) orders. Using the ResNet-50 trained on the ImageNet, we demonstrate the proposed fidelity estimation of four popular post-hoc interpretability methods.
comment: 26 pages, 9 figures
♻ ☆ TPC: Transformation-Specific Smoothing for Point Cloud Models ICML 2022
Point cloud models with neural network architectures have achieved great success and have been widely used in safety-critical applications, such as Lidar-based recognition systems in autonomous vehicles. However, such models are shown vulnerable to adversarial attacks which aim to apply stealthy semantic transformations such as rotation and tapering to mislead model predictions. In this paper, we propose a transformation-specific smoothing framework TPC, which provides tight and scalable robustness guarantees for point cloud models against semantic transformation attacks. We first categorize common 3D transformations into three categories: additive (e.g., shearing), composable (e.g., rotation), and indirectly composable (e.g., tapering), and we present generic robustness certification strategies for all categories respectively. We then specify unique certification protocols for a range of specific semantic transformations and their compositions. Extensive experiments on several common 3D transformations show that TPC significantly outperforms the state of the art. For example, our framework boosts the certified accuracy against twisting transformation along z-axis (within 20$^\circ$) from 20.3$\%$ to 83.8$\%$. Codes and models are available at https://github.com/Qianhewu/Point-Cloud-Smoothing.
comment: Accepted as a conference paper at ICML 2022
♻ ☆ Is it enough to optimize CNN architectures on ImageNet?
Classification performance based on ImageNet is the de-facto standard metric for CNN development. In this work we challenge the notion that CNN architecture design solely based on ImageNet leads to generally effective convolutional neural network (CNN) architectures that perform well on a diverse set of datasets and application domains. To this end, we investigate and ultimately improve ImageNet as a basis for deriving such architectures. We conduct an extensive empirical study for which we train $500$ CNN architectures, sampled from the broad AnyNetX design space, on ImageNet as well as $8$ additional well known image classification benchmark datasets from a diverse array of application domains. We observe that the performances of the architectures are highly dataset dependent. Some datasets even exhibit a negative error correlation with ImageNet across all architectures. We show how to significantly increase these correlations by utilizing ImageNet subsets restricted to fewer classes. These contributions can have a profound impact on the way we design future CNN architectures and help alleviate the tilt we see currently in our community with respect to over-reliance on one dataset.
♻ ☆ Provably Uncertainty-Guided Universal Domain Adaptation
Universal domain adaptation (UniDA) aims to transfer the knowledge of common classes from source domain to target domain without any prior knowledge on the label set, which requires to distinguish the unknown samples from the known ones in the target domain. A main challenge of UniDA is that the unequal label spaces of both domains causes the misalignment between two domains.To address the above challenging problems, we propose a new uncertainty-guided UniDA framework. Firstly, we introduce an empirical estimation of the probability of a target sample belonging to the unknown class with exploiting the distribution of target samples. Then, based on the estimation, we propose a novel neighbors searching method in the linear subspace with a $\delta$-filter to estimate the uncertainty score of a target sample and discover unknown samples. It fully utilizes the relationship between a target sample and its neighbors in source domain to avoid the influence of domain misalignment. Secondly, this paper well balances the confidence of predictions for both known and unknown samples through an uncertainty-guided margin loss based on the predictions of discovered unknown samples, which can reduce the gap between intra-class variance of known classes with respect to the unknown class. Finally, experiments on three public datasets demonstrate that our method significantly outperforms existing state-of-the-art methods.
comment: 13 pages. arXiv admin note: text overlap with arXiv:2207.09280
♻ ☆ Understanding and Improving Visual Prompting: A Label-Mapping Perspective CVPR 2023
We revisit and advance visual prompting (VP), an input prompting technique for vision tasks. VP can reprogram a fixed, pre-trained source model to accomplish downstream tasks in the target domain by simply incorporating universal prompts (in terms of input perturbation patterns) into downstream data points. Yet, it remains elusive why VP stays effective even given a ruleless label mapping (LM) between the source classes and the target classes. Inspired by the above, we ask: How is LM interrelated with VP? And how to exploit such a relationship to improve its accuracy on target tasks? We peer into the influence of LM on VP and provide an affirmative answer that a better 'quality' of LM (assessed by mapping precision and explanation) can consistently improve the effectiveness of VP. This is in contrast to the prior art where the factor of LM was missing. To optimize LM, we propose a new VP framework, termed ILM-VP (iterative label mapping-based visual prompting), which automatically re-maps the source labels to the target labels and progressively improves the target task accuracy of VP. Further, when using a contrastive language-image pretrained (CLIP) model, we propose to integrate an LM process to assist the text prompt selection of CLIP and to improve the target task accuracy. Extensive experiments demonstrate that our proposal significantly outperforms state-of-the-art VP methods. As highlighted below, we show that when reprogramming an ImageNet-pretrained ResNet-18 to 13 target tasks, our method outperforms baselines by a substantial margin, e.g., 7.9% and 6.7% accuracy improvements in transfer learning to the target Flowers102 and CIFAR100 datasets. Besides, our proposal on CLIP-based VP provides 13.7% and 7.1% accuracy improvements on Flowers102 and DTD respectively. Our code is available at https://github.com/OPTML-Group/ILM-VP.
comment: CVPR 2023
♻ ☆ DeepSolo: Let Transformer Decoder with Explicit Points Solo for Text Spotting CVPR 2023
End-to-end text spotting aims to integrate scene text detection and recognition into a unified framework. Dealing with the relationship between the two sub-tasks plays a pivotal role in designing effective spotters. Although Transformer-based methods eliminate the heuristic post-processing, they still suffer from the synergy issue between the sub-tasks and low training efficiency. In this paper, we present DeepSolo, a simple DETR-like baseline that lets a single Decoder with Explicit Points Solo for text detection and recognition simultaneously. Technically, for each text instance, we represent the character sequence as ordered points and model them with learnable explicit point queries. After passing a single decoder, the point queries have encoded requisite text semantics and locations, thus can be further decoded to the center line, boundary, script, and confidence of text via very simple prediction heads in parallel. Besides, we also introduce a text-matching criterion to deliver more accurate supervisory signals, thus enabling more efficient training. Quantitative experiments on public benchmarks demonstrate that DeepSolo outperforms previous state-of-the-art methods and achieves better training efficiency. In addition, DeepSolo is also compatible with line annotations, which require much less annotation cost than polygons. The code is available at https://github.com/ViTAE-Transformer/DeepSolo.
comment: CVPR 2023
♻ ☆ CLIP the Gap: A Single Domain Generalization Approach for Object Detection
Single Domain Generalization (SDG) tackles the problem of training a model on a single source domain so that it generalizes to any unseen target domain. While this has been well studied for image classification, the literature on SDG object detection remains almost non-existent. To address the challenges of simultaneously learning robust object localization and representation, we propose to leverage a pre-trained vision-language model to introduce semantic domain concepts via textual prompts. We achieve this via a semantic augmentation strategy acting on the features extracted by the detector backbone, as well as a text-based classification loss. Our experiments evidence the benefits of our approach, outperforming by 10% the only existing SDG object detection method, Single-DGOD [49], on their own diverse weather-driving benchmark.
♻ ☆ Leveraging Different Learning Styles for Improved Knowledge Distillation
Learning style refers to a type of training mechanism adopted by an individual to gain new knowledge. As suggested by the VARK model, humans have different learning preferences like visual, auditory, etc., for acquiring and effectively processing information. Inspired by this concept, our work explores the idea of mixed information sharing with model compression in the context of Knowledge Distillation (KD) and Mutual Learning (ML). Unlike conventional techniques that share the same type of knowledge with all networks, we propose to train individual networks with different forms of information to enhance the learning process. We formulate a combined KD and ML framework with one teacher and two student networks that share or exchange information in the form of predictions and feature maps. Our comprehensive experiments with benchmark classification and segmentation datasets demonstrate that with 15% compression, the ensemble performance of networks trained with diverse forms of knowledge outperforms the conventional techniques both quantitatively and qualitatively.
♻ ☆ ROIFormer: Semantic-Aware Region of Interest Transformer for Efficient Self-Supervised Monocular Depth Estimation AAAI 2023
The exploration of mutual-benefit cross-domains has shown great potential toward accurate self-supervised depth estimation. In this work, we revisit feature fusion between depth and semantic information and propose an efficient local adaptive attention method for geometric aware representation enhancement. Instead of building global connections or deforming attention across the feature space without restraint, we bound the spatial interaction within a learnable region of interest. In particular, we leverage geometric cues from semantic information to learn local adaptive bounding boxes to guide unsupervised feature aggregation. The local areas preclude most irrelevant reference points from attention space, yielding more selective feature learning and faster convergence. We naturally extend the paradigm into a multi-head and hierarchic way to enable the information distillation in different semantic levels and improve the feature discriminative ability for fine-grained depth estimation. Extensive experiments on the KITTI dataset show that our proposed method establishes a new state-of-the-art in self-supervised monocular depth estimation task, demonstrating the effectiveness of our approach over former Transformer variants.
comment: Camera Ready for AAAI 2023
♻ ☆ Consistent Attack: Universal Adversarial Perturbation on Embodied Vision Navigation
Embodied agents in vision navigation coupled with deep neural networks have attracted increasing attention. However, deep neural networks have been shown vulnerable to malicious adversarial noises, which may potentially cause catastrophic failures in Embodied Vision Navigation. Among different adversarial noises, universal adversarial perturbations (UAP), i.e., a constant image-agnostic perturbation applied on every input frame of the agent, play a critical role in Embodied Vision Navigation since they are computation-efficient and application-practical during the attack. However, existing UAP methods ignore the system dynamics of Embodied Vision Navigation and might be sub-optimal. In order to extend UAP to the sequential decision setting, we formulate the disturbed environment under the universal noise $\delta$, as a $\delta$-disturbed Markov Decision Process ($\delta$-MDP). Based on the formulation, we analyze the properties of $\delta$-MDP and propose two novel Consistent Attack methods, named Reward UAP and Trajectory UAP, for attacking Embodied agents, which consider the dynamic of the MDP and calculate universal noises by estimating the disturbed distribution and the disturbed Q function. For various victim models, our Consistent Attack can cause a significant drop in their performance in the PointGoal task in Habitat with different datasets and different scenes. Extensive experimental results indicate that there exist serious potential risks for applying Embodied Vision Navigation methods to the real world.
♻ ☆ Red Blood Cell Segmentation with Overlapping Cell Separation and Classification on Imbalanced Dataset
Automated red blood cell (RBC) classification on blood smear images helps hematologists to analyze RBC lab results in a reduced time and cost. However, overlapping cells can cause incorrect predicted results, and so they have to be separated into multiple single RBCs before classifying. To classify multiple classes with deep learning, imbalance problems are common in medical imaging because normal samples are always higher than rare disease samples. This paper presents a new method to segment and classify RBCs from blood smear images, specifically to tackle cell overlapping and data imbalance problems. Focusing on overlapping cell separation, our segmentation process first estimates ellipses to represent RBCs. The method detects the concave points and then finds the ellipses using directed ellipse fitting. The accuracy from 20 blood smear images was 0.889. Classification requires balanced training datasets. However, some RBC types are rare. The imbalance ratio of this dataset was 34.538 for 12 RBC classes from 20,875 individual RBC samples. The use of machine learning for RBC classification with an imbalanced dataset is hence more challenging than many other applications. We analyzed techniques to deal with this problem. The best accuracy and F1-score were 0.921 and 0.8679, respectively, using EfficientNet-B1 with augmentation. Experimental results showed that the weight balancing technique with augmentation had the potential to deal with imbalance problems by improving the F1-score on minority classes, while data augmentation significantly improved the overall classification performance.
comment: This work has been submitted to Intelligent Systems with Applications (ISWA) for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessible
♻ ☆ Texture-guided Saliency Distilling for Unsupervised Salient Object Detection CVPR 2023
Deep Learning-based Unsupervised Salient Object Detection (USOD) mainly relies on the noisy saliency pseudo labels that have been generated from traditional handcraft methods or pre-trained networks. To cope with the noisy labels problem, a class of methods focus on only easy samples with reliable labels but ignore valuable knowledge in hard samples. In this paper, we propose a novel USOD method to mine rich and accurate saliency knowledge from both easy and hard samples. First, we propose a Confidence-aware Saliency Distilling (CSD) strategy that scores samples conditioned on samples' confidences, which guides the model to distill saliency knowledge from easy samples to hard samples progressively. Second, we propose a Boundary-aware Texture Matching (BTM) strategy to refine the boundaries of noisy labels by matching the textures around the predicted boundary. Extensive experiments on RGB, RGB-D, RGB-T, and video SOD benchmarks prove that our method achieves state-of-the-art USOD performance.
comment: 8 pages, accepted to CVPR 2023
♻ ☆ GenDexGrasp: Generalizable Dexterous Grasping ICRA 2023
Generating dexterous grasping has been a long-standing and challenging robotic task. Despite recent progress, existing methods primarily suffer from two issues. First, most prior arts focus on a specific type of robot hand, lacking the generalizable capability of handling unseen ones. Second, prior arts oftentimes fail to rapidly generate diverse grasps with a high success rate. To jointly tackle these challenges with a unified solution, we propose GenDexGrasp, a novel hand-agnostic grasping algorithm for generalizable grasping. GenDexGrasp is trained on our proposed large-scale multi-hand grasping dataset MultiDex synthesized with force closure optimization. By leveraging the contact map as a hand-agnostic intermediate representation, GenDexGrasp efficiently generates diverse and plausible grasping poses with a high success rate and can transfer among diverse multi-fingered robotic hands. Compared with previous methods, GenDexGrasp achieves a three-way trade-off among success rate, inference speed, and diversity. Code is available at https://github.com/tengyu-liu/GenDexGrasp.
comment: Accepted to ICRA 2023 (camera-ready version)
♻ ☆ PI-Trans: Parallel-ConvMLP and Implicit-Transformation Based GAN for Cross-View Image Translation
For semantic-guided cross-view image translation, it is crucial to learn where to sample pixels from the source view image and where to reallocate them guided by the target view semantic map, especially when there is little overlap or drastic view difference between the source and target images. Hence, one not only needs to encode the long-range dependencies among pixels in both the source view image and target view semantic map but also needs to translate these learned dependencies. To this end, we propose a novel generative adversarial network, PI-Trans, which mainly consists of a novel Parallel-ConvMLP module and an Implicit Transformation module at multiple semantic levels. Extensive experimental results show that PI-Trans achieves the best qualitative and quantitative performance by a large margin compared to the state-of-the-art methods on two challenging datasets. The source code is available at https://github.com/Amazingren/PI-Trans.
comment: 5 pages, 5 figures
♻ ☆ Fine-grained Retrieval Prompt Tuning AAAI 2023
Fine-grained object retrieval aims to learn discriminative representation to retrieve visually similar objects. However, existing top-performing works usually impose pairwise similarities on the semantic embedding spaces or design a localization sub-network to continually fine-tune the entire model in limited data scenarios, thus resulting in convergence to suboptimal solutions. In this paper, we develop Fine-grained Retrieval Prompt Tuning (FRPT), which steers a frozen pre-trained model to perform the fine-grained retrieval task from the perspectives of sample prompting and feature adaptation. Specifically, FRPT only needs to learn fewer parameters in the prompt and adaptation instead of fine-tuning the entire model, thus solving the issue of convergence to suboptimal solutions caused by fine-tuning the entire model. Technically, a discriminative perturbation prompt (DPP) is introduced and deemed as a sample prompting process, which amplifies and even exaggerates some discriminative elements contributing to category prediction via a content-aware inhomogeneous sampling operation. In this way, DPP can make the fine-grained retrieval task aided by the perturbation prompts close to the solved task during the original pre-training. Thereby, it preserves the generalization and discrimination of representation extracted from input samples. Besides, a category-specific awareness head is proposed and regarded as feature adaptation, which removes the species discrepancies in features extracted by the pre-trained model using category-guided instance normalization. And thus, it makes the optimized features only include the discrepancies among subcategories. Extensive experiments demonstrate that our FRPT with fewer learnable parameters achieves the state-of-the-art performance on three widely-used fine-grained datasets.
comment: Accepted by AAAI 2023
♻ ☆ Mask-FPAN: Semi-Supervised Face Parsing in the Wild With De-Occlusion and UV GAN
Fine-grained semantic segmentation of a person's face and head, including facial parts and head components, has progressed a great deal in recent years. However, it remains a challenging task, whereby considering ambiguous occlusions and large pose variations are particularly difficult. To overcome these difficulties, we propose a novel framework termed Mask-FPAN. It uses a de-occlusion module that learns to parse occluded faces in a semi-supervised way. In particular, face landmark localization, face occlusionstimations, and detected head poses are taken into account. A 3D morphable face model combined with the UV GAN improves the robustness of 2D face parsing. In addition, we introduce two new datasets named FaceOccMask-HQ and CelebAMaskOcc-HQ for face paring work. The proposed Mask-FPAN framework addresses the face parsing problem in the wild and shows significant performance improvements with MIOU from 0.7353 to 0.9013 compared to the state-of-the-art on challenging face datasets.
comment: renew later
♻ ☆ Multi-Order Networks for Action Unit Detection
Action Units (AU) are muscular activations used to describe facial expressions. Therefore accurate AU recognition unlocks unbiaised face representation which can improve face-based affective computing applications. From a learning standpoint AU detection is a multi-task problem with strong inter-task dependencies. To solve such problem, most approaches either rely on weight sharing, or add explicit dependency modelling by decomposing the joint task distribution using Bayes chain rule. If the latter strategy yields comprehensive inter-task relationships modelling, it requires imposing an arbitrary order into an unordered task set. Crucially, this ordering choice has been identified as a source of performance variations. In this paper, we present Multi-Order Network (MONET), a multi-task method with joint task order optimization. MONET uses a differentiable order selection to jointly learn task-wise modules with their optimal chaining order. Furthermore, we introduce warmup and order dropout to enhance order selection by encouraging order exploration. Experimentally, we first demonstrate MONET capacity to retrieve the optimal order in a toy environment. Second, we validate MONET architecture by showing that MONET outperforms existing multi-task baselines on multiple attribute detection problems chosen for their wide range of dependency settings. More importantly, we demonstrate that MONET significantly extends state-of-the-art performance in AU detection.
♻ ☆ Bilateral-Fuser: A Novel Multi-cue Fusion Architecture with Anatomical-aware Tokens for Fovea Localization
Accurate localization of the fovea is a crucial initial step in analyzing retinal diseases since it helps prevent irreversible vision loss. Although current deep learning-based methods achieve better performance than traditional methods, they still face challenges such as inadequate utilization of anatomical landmarks, sensitivity to diseased retinal images, and various image conditions. In this paper, we propose a novel transformer-based architecture (Bilateral-Fuser) for multi-cue fusion. The Bilateral-Fuser explicitly incorporates long-range connections and global features using retina and vessel distributions to achieve robust fovea localization. We introduce a spatial attention mechanism in the dual-stream encoder to extract and fuse self-learned anatomical information. This design focuses more on features distributed along blood vessels and significantly reduces computational costs by reducing token numbers. Our comprehensive experiments demonstrate that the proposed architecture achieves state-of-the-art performance on two public datasets and one large-scale private dataset. Moreover, we show that the Bilateral-Fuser is more robust on both normal and diseased retina images and has better generalization capacity in cross-dataset experiments.
comment: This paper is prepared for IEEE Transactions on Medical Imaging
♻ ☆ Unproportional mosaicing
Data shift is a gap between data distribution used for training and data distribution encountered in the real-world. Data augmentations help narrow the gap by generating new data samples, increasing data variability, and data space coverage. We present a new data augmentation: Unproportional mosaicing (Unprop). Our augmentation randomly splits an image into various-sized blocks and swaps its content (pixels) while maintaining block sizes. Our method achieves a lower error rate when combined with other state-of-the-art augmentations.
♻ ☆ NAR-Former: Neural Architecture Representation Learning towards Holistic Attributes Prediction CVPR
With the wide and deep adoption of deep learning models in real applications, there is an increasing need to model and learn the representations of the neural networks themselves. These models can be used to estimate attributes of different neural network architectures such as the accuracy and latency, without running the actual training or inference tasks. In this paper, we propose a neural architecture representation model that can be used to estimate these attributes holistically. Specifically, we first propose a simple and effective tokenizer to encode both the operation and topology information of a neural network into a single sequence. Then, we design a multi-stage fusion transformer to build a compact vector representation from the converted sequence. For efficient model training, we further propose an information flow consistency augmentation and correspondingly design an architecture consistency loss, which brings more benefits with less augmentation samples compared with previous random augmentation strategies. Experiment results on NAS-Bench-101, NAS-Bench-201, DARTS search space and NNLQP show that our proposed framework can be used to predict the aforementioned latency and accuracy attributes of both cell architectures and whole deep neural networks, and achieves promising performance.
comment: 9 pages, 4 figures, 7 tables. Accepted by Computer Vision and Pattern Recognition (CVPR)2023. The code will be released soon
♻ ☆ Data Leakage and Evaluation Issues in Micro-Expression Analysis
Micro-expressions have drawn increasing interest lately due to various potential applications. The task is, however, difficult as it incorporates many challenges from the fields of computer vision, machine learning and emotional sciences. Due to the spontaneous and subtle characteristics of micro-expressions, the available training and testing data are limited, which make evaluation complex. We show that data leakage and fragmented evaluation protocols are issues among the micro-expression literature. We find that fixing data leaks can drastically reduce model performance, in some cases even making the models perform similarly to a random classifier. To this end, we go through common pitfalls, propose a new standardized evaluation protocol using facial action units with over 2000 micro-expression samples, and provide an open source library that implements the evaluation protocols in a standardized manner. Code is publicly available in \url{https://github.com/tvaranka/meb}.
♻ ☆ HybridCap: Inertia-aid Monocular Capture of Challenging Human Motions
Monocular 3D motion capture (mocap) is beneficial to many applications. The use of a single camera, however, often fails to handle occlusions of different body parts and hence it is limited to capture relatively simple movements. We present a light-weight, hybrid mocap technique called HybridCap that augments the camera with only 4 Inertial Measurement Units (IMUs) in a learning-and-optimization framework. We first employ a weakly-supervised and hierarchical motion inference module based on cooperative Gated Recurrent Unit (GRU) blocks that serve as limb, body and root trackers as well as an inverse kinematics solver. Our network effectively narrows the search space of plausible motions via coarse-to-fine pose estimation and manages to tackle challenging movements with high efficiency. We further develop a hybrid optimization scheme that combines inertial feedback and visual cues to improve tracking accuracy. Extensive experiments on various datasets demonstrate HybridCap can robustly handle challenging movements ranging from fitness actions to Latin dance. It also achieves real-time performance up to 60 fps with state-of-the-art accuracy.
♻ ☆ Spectral Geometric Verification: Re-Ranking Point Cloud Retrieval for Metric Localization
In large-scale metric localization, an incorrect result during retrieval will lead to an incorrect pose estimate or loop closure. Re-ranking methods propose to take into account all the top retrieval candidates and re-order them to increase the likelihood of the top candidate being correct. However, state-of-the-art re-ranking methods are inefficient when re-ranking many potential candidates due to their need for resource intensive point cloud registration between the query and each candidate. In this work, we propose an efficient spectral method for geometric verification (named SpectralGV) that does not require registration. We demonstrate how the optimal inter-cluster score of the correspondence compatibility graph of two point clouds represents a robust fitness score measuring their spatial consistency. This score takes into account the subtle geometric differences between structurally similar point clouds and therefore can be used to identify the correct candidate among potential matches retrieved by global similarity search. SpectralGV is deterministic, robust to outlier correspondences, and can be computed in parallel for all potential candidates. We conduct extensive experiments on 5 large-scale datasets to demonstrate that SpectralGV outperforms other state-of-the-art re-ranking methods and show that it consistently improves the recall and pose estimation of 3 state-of-the-art metric localization architectures while having a negligible effect on their runtime. The open-source implementation and trained models are available at: https://github.com/csiro-robotics/SpectralGV.
comment: Accepted for publication in IEEE RA-L (2023)
♻ ☆ Focus On Details: Online Multi-object Tracking with Diverse Fine-grained Representation CVPR2023
Discriminative representation is essential to keep a unique identifier for each target in Multiple object tracking (MOT). Some recent MOT methods extract features of the bounding box region or the center point as identity embeddings. However, when targets are occluded, these coarse-grained global representations become unreliable. To this end, we propose exploring diverse fine-grained representation, which describes appearance comprehensively from global and local perspectives. This fine-grained representation requires high feature resolution and precise semantic information. To effectively alleviate the semantic misalignment caused by indiscriminate contextual information aggregation, Flow Alignment FPN (FAFPN) is proposed for multi-scale feature alignment aggregation. It generates semantic flow among feature maps from different resolutions to transform their pixel positions. Furthermore, we present a Multi-head Part Mask Generator (MPMG) to extract fine-grained representation based on the aligned feature maps. Multiple parallel branches of MPMG allow it to focus on different parts of targets to generate local masks without label supervision. The diverse details in target masks facilitate fine-grained representation. Eventually, benefiting from a Shuffle-Group Sampling (SGS) training strategy with positive and negative samples balanced, we achieve state-of-the-art performance on MOT17 and MOT20 test sets. Even on DanceTrack, where the appearance of targets is extremely similar, our method significantly outperforms ByteTrack by 5.0% on HOTA and 5.6% on IDF1. Extensive experiments have proved that diverse fine-grained representation makes Re-ID great again in MOT.
comment: CVPR2023
♻ ☆ Can CNNs Be More Robust Than Transformers? ICLR2023
The recent success of Vision Transformers is shaking the long dominance of Convolutional Neural Networks (CNNs) in image recognition for a decade. Specifically, in terms of robustness on out-of-distribution samples, recent research finds that Transformers are inherently more robust than CNNs, regardless of different training setups. Moreover, it is believed that such superiority of Transformers should largely be credited to their self-attention-like architectures per se. In this paper, we question that belief by closely examining the design of Transformers. Our findings lead to three highly effective architecture designs for boosting robustness, yet simple enough to be implemented in several lines of code, namely a) patchifying input images, b) enlarging kernel size, and c) reducing activation layers and normalization layers. Bringing these components together, we are able to build pure CNN architectures without any attention-like operations that are as robust as, or even more robust than, Transformers. We hope this work can help the community better understand the design of robust neural architectures. The code is publicly available at https://github.com/UCSC-VLAA/RobustCNN.
comment: ICLR2023. Code is available at https://github.com/UCSC-VLAA/RobustCNN
♻ ☆ Vox-Fusion: Dense Tracking and Mapping with Voxel-based Neural Implicit Representation
In this work, we present a dense tracking and mapping system named Vox-Fusion, which seamlessly fuses neural implicit representations with traditional volumetric fusion methods. Our approach is inspired by the recently developed implicit mapping and positioning system and further extends the idea so that it can be freely applied to practical scenarios. Specifically, we leverage a voxel-based neural implicit surface representation to encode and optimize the scene inside each voxel. Furthermore, we adopt an octree-based structure to divide the scene and support dynamic expansion, enabling our system to track and map arbitrary scenes without knowing the environment like in previous works. Moreover, we proposed a high-performance multi-process framework to speed up the method, thus supporting some applications that require real-time performance. The evaluation results show that our methods can achieve better accuracy and completeness than previous methods. We also show that our Vox-Fusion can be used in augmented reality and virtual reality applications. Our source code is publicly available at https://github.com/zju3dv/Vox-Fusion.
♻ ☆ AISFormer: Amodal Instance Segmentation with Transformer BMVC2022
Amodal Instance Segmentation (AIS) aims to segment the region of both visible and possible occluded parts of an object instance. While Mask R-CNN-based AIS approaches have shown promising results, they are unable to model high-level features coherence due to the limited receptive field. The most recent transformer-based models show impressive performance on vision tasks, even better than Convolution Neural Networks (CNN). In this work, we present AISFormer, an AIS framework, with a Transformer-based mask head. AISFormer explicitly models the complex coherence between occluder, visible, amodal, and invisible masks within an object's regions of interest by treating them as learnable queries. Specifically, AISFormer contains four modules: (i) feature encoding: extract ROI and learn both short-range and long-range visual features. (ii) mask transformer decoding: generate the occluder, visible, and amodal mask query embeddings by a transformer decoder (iii) invisible mask embedding: model the coherence between the amodal and visible masks, and (iv) mask predicting: estimate output masks including occluder, visible, amodal and invisible. We conduct extensive experiments and ablation studies on three challenging benchmarks i.e. KINS, D2SA, and COCOA-cls to evaluate the effectiveness of AISFormer. The code is available at: https://github.com/UARK-AICV/AISFormer
comment: Accepted to BMVC2022
♻ ☆ DRKF: Distilled Rotated Kernel Fusion for Efficient Rotation Invariant Descriptors in Local Feature Matching
The performance of local feature descriptors degrades in the presence of large rotation variations. To address this issue, we present an efficient approach to learning rotation invariant descriptors. Specifically, we propose Rotated Kernel Fusion (RKF) which imposes rotations on each convolution kernel and improves the inherent nature of CNN. Since RKF can be processed by the subsequent re-parameterization, no extra computational costs will be introduced in the inference stage. Moreover, we present Multi-oriented Feature Aggregation (MOFA) which ensembles features extracted from multiple rotated versions of input images and can provide auxiliary information for the training of RKF by leveraging the knowledge distillation strategy. We refer to the distilled RKF model as DRKF. Besides the evaluation on a rotation-augmented version of the public dataset HPatches, we also contribute a new dataset named DiverseBEV which consists of bird's eye view images with large viewpoint changes and camera rotations. Extensive experiments show that our method can outperform other state-of-the-art techniques when exposed to large rotation variations.
comment: 8 pages, 7 figures
♻ ☆ EcoTTA: Memory-Efficient Continual Test-time Adaptation via Self-distilled Regularization CVPR 2023
This paper presents a simple yet effective approach that improves continual test-time adaptation (TTA) in a memory-efficient manner. TTA may primarily be conducted on edge devices with limited memory, so reducing memory is crucial but has been overlooked in previous TTA studies. In addition, long-term adaptation often leads to catastrophic forgetting and error accumulation, which hinders applying TTA in real-world deployments. Our approach consists of two components to address these issues. First, we present lightweight meta networks that can adapt the frozen original networks to the target domain. This novel architecture minimizes memory consumption by decreasing the size of intermediate activations required for backpropagation. Second, our novel self-distilled regularization controls the output of the meta networks not to deviate significantly from the output of the frozen original networks, thereby preserving well-trained knowledge from the source domain. Without additional memory, this regularization prevents error accumulation and catastrophic forgetting, resulting in stable performance even in long-term test-time adaptation. We demonstrate that our simple yet effective strategy outperforms other state-of-the-art methods on various benchmarks for image classification and semantic segmentation tasks. Notably, our proposed method with ResNet-50 and WideResNet-40 takes 86% and 80% less memory than the recent state-of-the-art method, CoTTA.
comment: Accepted to CVPR 2023
♻ ☆ Need for Objective Task-based Evaluation of Deep Learning-Based Denoising Methods: A Study in the Context of Myocardial Perfusion SPECT
Artificial intelligence-based methods have generated substantial interest in nuclear medicine. An area of significant interest has been using deep-learning (DL)-based approaches for denoising images acquired with lower doses, shorter acquisition times, or both. Objective evaluation of these approaches is essential for clinical application. DL-based approaches for denoising nuclear-medicine images have typically been evaluated using fidelity-based figures of merit (FoMs) such as RMSE and SSIM. However, these images are acquired for clinical tasks and thus should be evaluated based on their performance in these tasks. Our objectives were to (1) investigate whether evaluation with these FoMs is consistent with objective clinical-task-based evaluation; (2) provide a theoretical analysis for determining the impact of denoising on signal-detection tasks; (3) demonstrate the utility of virtual clinical trials (VCTs) to evaluate DL-based methods. A VCT to evaluate a DL-based method for denoising myocardial perfusion SPECT (MPS) images was conducted. The impact of DL-based denoising was evaluated using fidelity-based FoMs and AUC, which quantified performance on detecting perfusion defects in MPS images as obtained using a model observer with anthropomorphic channels. Based on fidelity-based FoMs, denoising using the considered DL-based method led to significantly superior performance. However, based on ROC analysis, denoising did not improve, and in fact, often degraded detection-task performance. The results motivate the need for objective task-based evaluation of DL-based denoising approaches. Further, this study shows how VCTs provide a mechanism to conduct such evaluations using VCTs. Finally, our theoretical treatment reveals insights into the reasons for the limited performance of the denoising approach.
♻ ☆ An Interpretable MRI Reconstruction Network with Two-grid-cycle Correction and Geometric Prior Distillation
Although existing deep learning compressed-sensing-based Magnetic Resonance Imaging (CS-MRI) methods have achieved considerably impressive performance, explainability and generalizability continue to be challenging for such methods since the transition from mathematical analysis to network design not always natural enough, often most of them are not flexible enough to handle multi-sampling-ratio reconstruction assignments. {In this work, to tackle explainability and generalizability, we propose a unifying deep unfolding multi-sampling-ratio interpretable CS-MRI framework.} The combined approach offers more generalizability than previous works whereas deep learning gains explainability through a geometric prior module. Inspired by the multigrid algorithm, we first embed the CS-MRI-based optimization algorithm into correction-distillation scheme that consists of three ingredients: pre-relaxation module, correction module and geometric prior distillation module. Furthermore, we employ a condition module to learn adaptively step-length and noise level, which enables the proposed framework to jointly train multi-ratio tasks through a single model. { The proposed model not only compensates for the lost contextual information of reconstructed image which is refined from low frequency error in geometric characteristic k-space}, but also integrates the theoretical guarantee of model-based methods and the superior reconstruction performances of deep learning-based methods. Therefore, it can give us a novel perspective to design biomedical imaging networks. { Numerical experiments show that our framework outperforms state-of-the-art methods in terms of qualitative and quantitative evaluations.} {Our method achieves 3.18 dB improvement at low CS ratio 10\% and average 1.42 dB improvement over other comparison methods on brain dataset using Cartesian sampling mask.
comment: 14 pages, accepted to Biomedical Signal Processing and Control,March, 2023
♻ ☆ CAT: LoCalization and IdentificAtion Cascade Detection Transformer for Open-World Object Detection
Open-world object detection (OWOD), as a more general and challenging goal, requires the model trained from data on known objects to detect both known and unknown objects and incrementally learn to identify these unknown objects. The existing works which employ standard detection framework and fixed pseudo-labelling mechanism (PLM) have the following problems: (i) The inclusion of detecting unknown objects substantially reduces the model's ability to detect known ones. (ii) The PLM does not adequately utilize the priori knowledge of inputs. (iii) The fixed selection manner of PLM cannot guarantee that the model is trained in the right direction. We observe that humans subconsciously prefer to focus on all foreground objects and then identify each one in detail, rather than localize and identify a single object simultaneously, for alleviating the confusion. This motivates us to propose a novel solution called CAT: LoCalization and IdentificAtion Cascade Detection Transformer which decouples the detection process via the shared decoder in the cascade decoding way. In the meanwhile, we propose the self-adaptive pseudo-labelling mechanism which combines the model-driven with input-driven PLM and self-adaptively generates robust pseudo-labels for unknown objects, significantly improving the ability of CAT to retrieve unknown objects. Comprehensive experiments on two benchmark datasets, i.e., MS-COCO and PASCAL VOC, show that our model outperforms the state-of-the-art in terms of all metrics in the task of OWOD, incremental object detection (IOD) and open-set detection.
♻ ☆ Backdoor Defense via Suppressing Model Shortcuts ICASSP 2023
Recent studies have demonstrated that deep neural networks (DNNs) are vulnerable to backdoor attacks during the training process. Specifically, the adversaries intend to embed hidden backdoors in DNNs so that malicious model predictions can be activated through pre-defined trigger patterns. In this paper, we explore the backdoor mechanism from the angle of the model structure. We select the skip connection for discussions, inspired by the understanding that it helps the learning of model `shortcuts' where backdoor triggers are usually easier to be learned. Specifically, we demonstrate that the attack success rate (ASR) decreases significantly when reducing the outputs of some key skip connections. Based on this observation, we design a simple yet effective backdoor removal method by suppressing the skip connections in critical layers selected by our method. We also implement fine-tuning on these layers to recover high benign accuracy and to further reduce ASR. Extensive experiments on benchmark datasets verify the effectiveness of our method.
comment: This paper is accepted by ICASSP 2023. 5 pages
♻ ☆ BATT: Backdoor Attack with Transformation-based Triggers ICASSP 2023
Deep neural networks (DNNs) are vulnerable to backdoor attacks. The backdoor adversaries intend to maliciously control the predictions of attacked DNNs by injecting hidden backdoors that can be activated by adversary-specified trigger patterns during the training process. One recent research revealed that most of the existing attacks failed in the real physical world since the trigger contained in the digitized test samples may be different from that of the one used for training. Accordingly, users can adopt spatial transformations as the image pre-processing to deactivate hidden backdoors. In this paper, we explore the previous findings from another side. We exploit classical spatial transformations (i.e. rotation and translation) with the specific parameter as trigger patterns to design a simple yet effective poisoning-based backdoor attack. For example, only images rotated to a particular angle can activate the embedded backdoor of attacked DNNs. Extensive experiments are conducted, verifying the effectiveness of our attack under both digital and physical settings and its resistance to existing backdoor defenses.
comment: This paper is accepted by ICASSP 2023. 5 pages
♻ ☆ Untargeted Backdoor Attack against Object Detection ICASSP 2023
Recent studies revealed that deep neural networks (DNNs) are exposed to backdoor threats when training with third-party resources (such as training samples or backbones). The backdoored model has promising performance in predicting benign samples, whereas its predictions can be maliciously manipulated by adversaries based on activating its backdoors with pre-defined trigger patterns. Currently, most of the existing backdoor attacks were conducted on the image classification under the targeted manner. In this paper, we reveal that these threats could also happen in object detection, posing threatening risks to many mission-critical applications ($e.g.$, pedestrian detection and intelligent surveillance systems). Specifically, we design a simple yet effective poison-only backdoor attack in an untargeted manner, based on task characteristics. We show that, once the backdoor is embedded into the target model by our attack, it can trick the model to lose detection of any object stamped with our trigger patterns. We conduct extensive experiments on the benchmark dataset, showing its effectiveness in both digital and physical-world settings and its resistance to potential defenses.
comment: This paper is accepted by ICASSP 2023. 5 pages
♻ ☆ Intrinsic Physical Concepts Discovery with Object-Centric Predictive Models CVPR
The ability to discover abstract physical concepts and understand how they work in the world through observing lies at the core of human intelligence. The acquisition of this ability is based on compositionally perceiving the environment in terms of objects and relations in an unsupervised manner. Recent approaches learn object-centric representations and capture visually observable concepts of objects, e.g., shape, size, and location. In this paper, we take a step forward and try to discover and represent intrinsic physical concepts such as mass and charge. We introduce the PHYsical Concepts Inference NEtwork (PHYCINE), a system that infers physical concepts in different abstract levels without supervision. The key insights underlining PHYCINE are two-fold, commonsense knowledge emerges with prediction, and physical concepts of different abstract levels should be reasoned in a bottom-up fashion. Empirical evaluation demonstrates that variables inferred by our system work in accordance with the properties of the corresponding physical concepts. We also show that object representations containing the discovered physical concepts variables could help achieve better performance in causal reasoning tasks, i.e., ComPhy.
comment: Accepted to Computer Vision and Pattern Recognition (CVPR)2023
♻ ☆ Ensembles of Compact, Region-specific & Regularized Spiking Neural Networks for Scalable Place Recognition ICRA
Spiking neural networks have significant potential utility in robotics due to their high energy efficiency on specialized hardware, but proof-of-concept implementations have not yet typically achieved competitive performance or capability with conventional approaches. In this paper, we tackle one of the key practical challenges of scalability by introducing a novel modular ensemble network approach, where compact, localized spiking networks each learn and are solely responsible for recognizing places in a local region of the environment only. This modular approach creates a highly scalable system. However, it comes with a high-performance cost where a lack of global regularization at deployment time leads to hyperactive neurons that erroneously respond to places outside their learned region. Our second contribution introduces a regularization approach that detects and removes these problematic hyperactive neurons during the initial environmental learning phase. We evaluate this new scalable modular system on benchmark localization datasets Nordland and Oxford RobotCar, with comparisons to standard techniques NetVLAD, DenseVLAD, and SAD, and a previous spiking neural network system. Our system substantially outperforms the previous SNN system on its small dataset, but also maintains performance on 27 times larger benchmark datasets where the operation of the previous system is computationally infeasible, and performs competitively with the conventional localization systems.
comment: 8 pages, 6 figures, accepted to the IEEE International Conference on Robotics and Automation (ICRA) 2023
♻ ☆ T360RRD: A dataset for 360 degree rotated rectangular box table detection
To address the problem of scarcity and high annotation costs of rotated image table detection datasets, this paper proposes a method for building a rotated image table detection dataset. Based on the ICDAR2019MTD modern table detection dataset, we refer to the annotation format of the DOTA dataset to create the TRR360D rotated table detection dataset. The training set contains 600 rotated images and 977 annotated instances, and the test set contains 240 rotated images and 499 annotated instances. The AP50(T<90) evaluation metric is defined, and this dataset is available for future researchers to study rotated table detection algorithms and promote the development of table detection technology. The TRR360D rotated table detection dataset was created by constraining the starting point and annotation direction, and is publicly available at https://github.com/vansin/TRR360D.
♻ ☆ Denoising diffusion models for out-of-distribution detection
Out-of-distribution detection is crucial to the safe deployment of machine learning systems. Currently, unsupervised out-of-distribution detection is dominated by generative-based approaches that make use of estimates of the likelihood or other measurements from a generative model. Reconstruction-based methods offer an alternative approach, in which a measure of reconstruction error is used to determine if a sample is out-of-distribution. However, reconstruction-based approaches are less favoured, as they require careful tuning of the model's information bottleneck - such as the size of the latent dimension - to produce good results. In this work, we exploit the view of denoising diffusion probabilistic models (DDPM) as denoising autoencoders where the bottleneck is controlled externally, by means of the amount of noise applied. We propose to use DDPMs to reconstruct an input that has been noised to a range of noise levels, and use the resulting multi-dimensional reconstruction error to classify out-of-distribution inputs. We validate our approach both on standard computer-vision datasets and on higher dimension medical datasets. Our approach outperforms not only reconstruction-based methods, but also state-of-the-art generative-based approaches.
♻ ☆ Benford's law: what does it say on adversarial images?
Convolutional neural networks (CNNs) are fragile to small perturbations in the input images. These networks are thus prone to malicious attacks that perturb the inputs to force a misclassification. Such slightly manipulated images aimed at deceiving the classifier are known as adversarial images. In this work, we investigate statistical differences between natural images and adversarial ones. More precisely, we show that employing a proper image transformation and for a class of adversarial attacks, the distribution of the leading digit of the pixels in adversarial images deviates from Benford's law. The stronger the attack, the more distant the resulting distribution is from Benford's law. Our analysis provides a detailed investigation of this new approach that can serve as a basis for alternative adversarial example detection methods that do not need to modify the original CNN classifier neither work on the raw high-dimensional pixels as features to defend against attacks.
♻ ☆ Rethinking Persistent Homology for Visual Recognition ICML 2022
Persistent topological properties of an image serve as an additional descriptor providing an insight that might not be discovered by traditional neural networks. The existing research in this area focuses primarily on efficiently integrating topological properties of the data in the learning process in order to enhance the performance. However, there is no existing study to demonstrate all possible scenarios where introducing topological properties can boost or harm the performance. This paper performs a detailed analysis of the effectiveness of topological properties for image classification in various training scenarios, defined by: the number of training samples, the complexity of the training data and the complexity of the backbone network. We identify the scenarios that benefit the most from topological features, e.g., training simple networks on small datasets. Additionally, we discuss the problem of topological consistency of the datasets which is one of the major bottlenecks for using topological features for classification. We further demonstrate how the topological inconsistency can harm the performance for certain scenarios.
comment: ICML 2022 Workshop on Topology, Algebra, and Geometry in Machine Learning
♻ ☆ Semi-supervised 3D Object Detection via Temporal Graph Neural Networks 3DV 2021
3D object detection plays an important role in autonomous driving and other robotics applications. However, these detectors usually require training on large amounts of annotated data that is expensive and time-consuming to collect. Instead, we propose leveraging large amounts of unlabeled point cloud videos by semi-supervised learning of 3D object detectors via temporal graph neural networks. Our insight is that temporal smoothing can create more accurate detection results on unlabeled data, and these smoothed detections can then be used to retrain the detector. We learn to perform this temporal reasoning with a graph neural network, where edges represent the relationship between candidate detections in different time frames. After semi-supervised learning, our method achieves state-of-the-art detection performance on the challenging nuScenes and H3D benchmarks, compared to baselines trained on the same amount of labeled data. Project and code are released at https://www.jianrenw.com/SOD-TGNN/.
comment: 3DV 2021
♻ ☆ Bridging the Gap to Real-World Object-Centric Learning ICLR 2023
Humans naturally decompose their environment into entities at the appropriate level of abstraction to act in the world. Allowing machine learning algorithms to derive this decomposition in an unsupervised way has become an important line of research. However, current methods are restricted to simulated data or require additional information in the form of motion or depth in order to successfully discover objects. In this work, we overcome this limitation by showing that reconstructing features from models trained in a self-supervised manner is a sufficient training signal for object-centric representations to arise in a fully unsupervised way. Our approach, DINOSAUR, significantly out-performs existing image-based object-centric learning models on simulated data and is the first unsupervised object-centric model that scales to real-world datasets such as COCO and PASCAL VOC. DINOSAUR is conceptually simple and shows competitive performance compared to more involved pipelines from the computer vision literature.
comment: ICLR 2023 camera-ready version
♻ ☆ Bag of Tricks for Long-Tail Visual Recognition of Animal Species in Camera-Trap Images
Camera traps are a method for monitoring wildlife and they collect a large number of pictures. The number of images collected of each species usually follows a long-tail distribution, i.e., a few classes have a large number of instances, while a lot of species have just a small percentage. Although in most cases these rare species are the ones of interest to ecologists, they are often neglected when using deep-learning models because these models require a large number of images for the training. In this work, a simple and effective framework called Square-Root Sampling Branch (SSB) is proposed, which combines two classification branches that are trained using square-root sampling and instance sampling to improve long-tail visual recognition, and this is compared to state-of-the-art methods for handling this task: square-root sampling, class-balanced focal loss, and balanced group softmax. To achieve a more general conclusion, the methods for handling long-tail visual recognition were systematically evaluated in four families of computer vision models (ResNet, MobileNetV3, EfficientNetV2, and Swin Transformer) and four camera-trap datasets with different characteristics. Initially, a robust baseline with the most recent training tricks was prepared and, then, the methods for improving long-tail recognition were applied. Our experiments show that square-root sampling was the method that most improved the performance for minority classes by around 15%; however, this was at the cost of reducing the majority classes' accuracy by at least 3%. Our proposed framework (SSB) demonstrated itself to be competitive with the other methods and achieved the best or the second-best results for most of the cases for the tail classes; but, unlike the square-root sampling, the loss in the performance of the head classes was minimal, thus achieving the best trade-off among all the evaluated methods.
♻ ☆ User Evaluation of Culture-to-Culture Image Translation with Generative Adversarial Nets
The article introduces the concept of image ``culturization," i.e., defined as the process of altering the ``brushstroke of cultural features" that make objects perceived as belonging to a given culture while preserving their functionalities. First, we defined a pipeline for translating objects' images from a source to a target cultural domain based on state-of-the-art Generative Adversarial Networks. Then, we gathered data through an online questionnaire to test four hypotheses concerning the impact of images belonging to different cultural domains on Italian participants. As expected, results depend on individual tastes and preferences: however, they align with our conjecture that some people, during the interaction with an intelligent system, will prefer to be shown images modified to match their cultural background.
comment: 40 pages (bibliography excluded), 5 figures, 6 Tables
♻ ☆ Local Neural Descriptor Fields: Locally Conditioned Object Representations for Manipulation ICRA 2023
A robot operating in a household environment will see a wide range of unique and unfamiliar objects. While a system could train on many of these, it is infeasible to predict all the objects a robot will see. In this paper, we present a method to generalize object manipulation skills acquired from a limited number of demonstrations, to novel objects from unseen shape categories. Our approach, Local Neural Descriptor Fields (L-NDF), utilizes neural descriptors defined on the local geometry of the object to effectively transfer manipulation demonstrations to novel objects at test time. In doing so, we leverage the local geometry shared between objects to produce a more general manipulation framework. We illustrate the efficacy of our approach in manipulating novel objects in novel poses -- both in simulation and in the real world.
comment: ICRA 2023, Project Page: https://elchun.github.io/lndf/
♻ ☆ Tuned Compositional Feature Replays for Efficient Stream Learning
Our brains extract durable, generalizable knowledge from transient experiences of the world. Artificial neural networks come nowhere close: when tasked with learning to classify objects by training on non-repeating video frames in temporal order (online stream learning), models that learn well from shuffled datasets catastrophically forget old knowledge upon learning new stimuli. We propose a new continual learning algorithm, Compositional Replay Using Memory Blocks (CRUMB), which mitigates forgetting by replaying feature maps reconstructed by recombining generic parts. Just as crumbs together form a loaf of bread, we concatenate trainable and re-usable "memory block" vectors to compositionally reconstruct feature map tensors in convolutional neural networks. CRUMB stores the indices of memory blocks used to reconstruct new stimuli, enabling replay of specific memories during later tasks. CRUMB's memory blocks are tuned to enhance replay: a single feature map stored, reconstructed, and replayed by CRUMB mitigates forgetting during video stream learning more effectively than an entire image, even though it occupies only 3.6% as much memory. We stress-tested CRUMB alongside 13 competing methods on 5 challenging datasets. To address the limited number of existing online stream learning datasets, we introduce 2 new benchmarks by adapting existing datasets for stream learning. With about 4% of the memory and 20% of the runtime, CRUMB mitigates catastrophic forgetting more effectively than the prior state-of-the-art. Our code is available at https://github.com/MorganBDT/crumb.git.
comment: Copyright 2023 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works
♻ ☆ Masked Video Distillation: Rethinking Masked Feature Modeling for Self-supervised Video Representation Learning CVPR 2023
Benefiting from masked visual modeling, self-supervised video representation learning has achieved remarkable progress. However, existing methods focus on learning representations from scratch through reconstructing low-level features like raw pixel RGB values. In this paper, we propose masked video distillation (MVD), a simple yet effective two-stage masked feature modeling framework for video representation learning: firstly we pretrain an image (or video) model by recovering low-level features of masked patches, then we use the resulting features as targets for masked feature modeling. For the choice of teacher models, we observe that students taught by video teachers perform better on temporally-heavy video tasks, while image teachers transfer stronger spatial representations for spatially-heavy video tasks. Visualization analysis also indicates different teachers produce different learned patterns for students. Motivated by this observation, we design a spatial-temporal co-teaching method for MVD. Specifically, we distill student models from both video teachers and image teachers by masked feature modeling. Extensive experimental results demonstrate that video transformers pretrained with spatial-temporal co-teaching outperform models distilled with a single teacher on a multitude of video datasets. Our MVD with vanilla ViT achieves state-of-the-art performance compared with previous supervised or self-supervised methods on several challenging video downstream tasks. For example, with the ViT-Large model, our MVD achieves 86.4% and 76.7% Top-1 accuracy on Kinetics-400 and Something-Something-v2, outperforming VideoMAE by 1.2% and 2.4% respectively. When a larger ViT-Huge model is adopted, MVD achieves the state-of-the-art performance with 77.3% Top-1 accuracy on Something-Something-v2 and 41.1 mAP on AVA v2.2. Code will be available at \url{https://github.com/ruiwang2021/mvd}.
comment: CVPR 2023, code will be available at https://github.com/ruiwang2021/mvd
♻ ☆ MultiViz: Towards Visualizing and Understanding Multimodal Models ICLR 2023
The promise of multimodal models for real-world applications has inspired research in visualizing and understanding their internal mechanics with the end goal of empowering stakeholders to visualize model behavior, perform model debugging, and promote trust in machine learning models. However, modern multimodal models are typically black-box neural networks, which makes it challenging to understand their internal mechanics. How can we visualize the internal modeling of multimodal interactions in these models? Our paper aims to fill this gap by proposing MultiViz, a method for analyzing the behavior of multimodal models by scaffolding the problem of interpretability into 4 stages: (1) unimodal importance: how each modality contributes towards downstream modeling and prediction, (2) cross-modal interactions: how different modalities relate with each other, (3) multimodal representations: how unimodal and cross-modal interactions are represented in decision-level features, and (4) multimodal prediction: how decision-level features are composed to make a prediction. MultiViz is designed to operate on diverse modalities, models, tasks, and research areas. Through experiments on 8 trained models across 6 real-world tasks, we show that the complementary stages in MultiViz together enable users to (1) simulate model predictions, (2) assign interpretable concepts to features, (3) perform error analysis on model misclassifications, and (4) use insights from error analysis to debug models. MultiViz is publicly available, will be regularly updated with new interpretation tools and metrics, and welcomes inputs from the community.
comment: ICLR 2023. Code available at: https://github.com/pliang279/MultiViz
♻ ☆ Joint cortical registration of geometry and function using semi-supervised learning
Brain surface-based image registration, an important component of brain image analysis, establishes spatial correspondence between cortical surfaces. Existing iterative and learning-based approaches focus on accurate registration of folding patterns of the cerebral cortex, and assume that geometry predicts function and thus functional areas will also be well aligned. However, structure/functional variability of anatomically corresponding areas across subjects has been widely reported. In this work, we introduce a learning-based cortical registration framework, JOSA, which jointly aligns folding patterns and functional maps while simultaneously learning an optimal atlas. We demonstrate that JOSA can substantially improve registration performance in both anatomical and functional domains over existing methods. By employing a semi-supervised training strategy, the proposed framework obviates the need for functional data during inference, enabling its use in broad neuroscientific domains where functional data may not be observed.
comment: B. Fischl and A. V. Dalca are co-senior authors with equal contributions
Information Retrieval
☆ Implementation of a noisy hyperlink removal system: A semantic and relatedness approach
As the volume of data on the web grows, the web structure graph, which is a graph representation of the web, continues to evolve. The structure of this graph has gradually shifted from content-based to non-content-based. Furthermore, spam data, such as noisy hyperlinks, in the web structure graph adversely affect the speed and efficiency of information retrieval and link mining algorithms. Previous works in this area have focused on removing noisy hyperlinks using structural and string approaches. However, these approaches may incorrectly remove useful links or be unable to detect noisy hyperlinks in certain circumstances. In this paper, a data collection of hyperlinks is initially constructed using an interactive crawler. The semantic and relatedness structure of the hyperlinks is then studied through semantic web approaches and tools such as the DBpedia ontology. Finally, the removal process of noisy hyperlinks is carried out using a reasoner on the DBpedia ontology. Our experiments demonstrate the accuracy and ability of semantic web technologies to remove noisy hyperlinks
☆ AmQA: Amharic Question Answering Dataset
Question Answering (QA) returns concise answers or answer lists from natural language text given a context document. Many resources go into curating QA datasets to advance robust models' development. There is a surge of QA datasets for languages like English, however, this is not true for Amharic. Amharic, the official language of Ethiopia, is the second most spoken Semitic language in the world. There is no published or publicly available Amharic QA dataset. Hence, to foster the research in Amharic QA, we present the first Amharic QA (AmQA) dataset. We crowdsourced 2628 question-answer pairs over 378 Wikipedia articles. Additionally, we run an XLMR Large-based baseline model to spark open-domain QA research interest. The best-performing baseline achieves an F-score of 69.58 and 71.74 in reader-retriever QA and reading comprehension settings respectively.
☆ LongEval-Retrieval: French-English Dynamic Test Collection for Continuous Web Search Evaluation
LongEval-Retrieval is a Web document retrieval benchmark that focuses on continuous retrieval evaluation. This test collection is intended to be used to study the temporal persistence of Information Retrieval systems and will be used as the test collection in the Longitudinal Evaluation of Model Performance Track (LongEval) at CLEF 2023. This benchmark simulates an evolving information system environment - such as the one a Web search engine operates in - where the document collection, the query distribution, and relevance all move continuously, while following the Cranfield paradigm for offline evaluation. To do that, we introduce the concept of a dynamic test collection that is composed of successive sub-collections each representing the state of an information system at a given time step. In LongEval-Retrieval, each sub-collection contains a set of queries, documents, and soft relevance assessments built from click models. The data comes from Qwant, a privacy-preserving Web search engine that primarily focuses on the French market. LongEval-Retrieval also provides a 'mirror' collection: it is initially constructed in the French language to benefit from the majority of Qwant's traffic, before being translated to English. This paper presents the creation process of LongEval-Retrieval and provides baseline runs and analysis.
☆ MABNet: Master Assistant Buddy Network with Hybrid Learning for Image Retrieval ICASSP
Image retrieval has garnered growing interest in recent times. The current approaches are either supervised or self-supervised. These methods do not exploit the benefits of hybrid learning using both supervision and self-supervision. We present a novel Master Assistant Buddy Network (MABNet) for image retrieval which incorporates both learning mechanisms. MABNet consists of master and assistant blocks, both learning independently through supervision and collectively via self-supervision. The master guides the assistant by providing its knowledge base as a reference for self-supervision and the assistant reports its knowledge back to the master by weight transfer. We perform extensive experiments on public datasets with and without post-processing.
comment: Accepted at International Conference on Acoustics, Speech, and Signal Processing (ICASSP) 2023
☆ Privacy-Preserving Fair Item Ranking
Users worldwide access massive amounts of curated data in the form of rankings on a daily basis. The societal impact of this ease of access has been studied and work has been done to propose and enforce various notions of fairness in rankings. Current computational methods for fair item ranking rely on disclosing user data to a centralized server, which gives rise to privacy concerns for the users. This work is the first to advance research at the conjunction of producer (item) fairness and consumer (user) privacy in rankings by exploring the incorporation of privacy-preserving techniques; specifically, differential privacy and secure multi-party computation. Our work extends the equity of amortized attention ranking mechanism to be privacy-preserving, and we evaluate its effects with respect to privacy, fairness, and ranking quality. Our results using real-world datasets show that we are able to effectively preserve the privacy of users and mitigate unfairness of items without making additional sacrifices to the quality of rankings in comparison to the ranking mechanism in the clear.
☆ Dual Feedback Attention Framework via Boundary-Aware Auxiliary and Progressive Semantic Optimization for Salient Object Detection in Optical Remote Sensing Imagery
Salient object detection in optical remote sensing image (ORSI-SOD) has gradually attracted attention thanks to the development of deep learning (DL) and salient object detection in natural scene image (NSI-SOD). However, NSI and ORSI are different in many aspects, such as large coverage, complex background, and large differences in target types and scales. Therefore, a new dedicated method is needed for ORSI-SOD. In addition, existing methods do not pay sufficient attention to the boundary of the object, and the completeness of the final saliency map still needs improvement. To address these issues, we propose a novel method called Dual Feedback Attention Framework via Boundary-Aware Auxiliary and Progressive Semantic Optimization (DFA-BASO). First, Boundary Protection Calibration (BPC) module is proposed to reduce the loss of edge position information during forward propagation and suppress noise in low-level features. Second, a Dual Feature Feedback Complementary (DFFC) module is proposed based on BPC module. It aggregates boundary-semantic dual features and provides effective feedback to coordinate features across different layers. Finally, a Strong Semantic Feedback Refinement (SSFR) module is proposed to obtain more complete saliency maps. This module further refines feature representation and eliminates feature differences through a unique feedback mechanism. Extensive experiments on two public datasets show that DFA-BASO outperforms 15 state-of-the-art methods. Furthermore, this paper strongly demonstrates the true contribution of DFA-BASO to ORSI-SOD by in-depth analysis of the visualization figure. All codes can be found at https://github.com/YUHsss/DFA-BASO.
☆ A Survey on Incremental Update for Neural Recommender Systems
Recommender Systems (RS) aim to provide personalized suggestions of items for users against consumer over-choice. Although extensive research has been conducted to address different aspects and challenges of RS, there still exists a gap between academic research and industrial applications. Specifically, most of the existing models still work in an offline manner, in which the recommender is trained on a large static training set and evaluated on a very restrictive testing set in a one-time process. RS will stay unchanged until the next batch retrain is performed. We frame such RS as Batch Update Recommender Systems (BURS). In reality, they have to face the challenges where RS are expected to be instantly updated with new data streaming in, and generate updated recommendations for current user activities based on the newly arrived data. We frame such RS as Incremental Update Recommender Systems (IURS). In this article, we offer a systematic survey of incremental update for neural recommender systems. We begin the survey by introducing key concepts and formulating the task of IURS. We then illustrate the challenges in IURS compared with traditional BURS. Afterwards, we detail the introduction of existing literature and evaluation issues. We conclude the survey by outlining some prominent open research issues in this area.
comment: 18 pages
☆ How Auditing Methodologies Can Impact Our Understanding of YouTube's Recommendation Systems
Data generated by audits of social media websites have formed the basis of our understanding of the biases presented in algorithmic content recommendation systems. As legislators around the world are beginning to consider regulating the algorithmic systems that drive online platforms, it is critical to ensure the correctness of these inferred biases. However, as we will show in this paper, doing so is a challenging task for a variety of reasons related to the complexity of configuration parameters associated with the audits that gather data from a specific platform. Focusing specifically on YouTube, we show that conducting audits to make inferences about YouTube's recommendation systems is more methodologically challenging than one might expect. There are many methodological decisions that need to be considered in order to obtain scientifically valid results, and each of these decisions incur costs. For example, should an auditor use (expensive to obtain) logged-in YouTube accounts while gathering recommendations from the algorithm to obtain more accurate inferences? We explore the impact of this and many other decisions and make some startling discoveries about the methodological choices that impact YouTube's recommendations. Taken all together, our research suggests auditing configuration compromises that YouTube auditors and researchers can use to reduce audit overhead, both economically and computationally, without sacrificing accuracy of their inferences. Similarly, we also identify several configuration parameters that have a significant impact on the accuracy of measured inferences and should be carefully considered.
♻ ☆ Learning to Recommend Using Non-Uniform Data
Learning user preferences for products based on their past purchases or reviews is at the cornerstone of modern recommendation engines. One complication in this learning task is that some users are more likely to purchase products or review them, and some products are more likely to be purchased or reviewed by the users. This non-uniform pattern degrades the power of many existing recommendation algorithms, as they assume that the observed data are sampled uniformly at random among user-product pairs. In addition, existing literature on modeling non-uniformity either assume user interests are independent of the products, or lack theoretical understanding. In this paper, we first model the user-product preferences as a partially observed matrix with non-uniform observation pattern. Next, building on the literature about low-rank matrix estimation, we introduce a new weighted trace-norm penalized regression to predict unobserved values of the matrix. We then prove an upper bound for the prediction error of our proposed approach. Our upper bound is a function of a number of parameters that are based on a certain weight matrix that depends on the joint distribution of users and products. Utilizing this observation, we introduce a new optimization problem to select a weight matrix that minimizes the upper bound on the prediction error. The final product is a new estimator, NU-Recommend, that outperforms existing methods in both synthetic and real datasets. Our approach aims at accurate predictions for all users while prioritizing fairness. To achieve this, we employ a bias-variance tradeoff mechanism that ensures good overall prediction performance without compromising the predictive accuracy for less active users.
♻ ☆ Complex QA and language models hybrid architectures, Survey
This paper provides a survey of the state of the art of hybrid language models architectures and strategies for "complex" question-answering (QA, CQA, CPS). Very large language models are good at leveraging public data on standard problems but once you want to tackle more specific complex questions or problems you may need specific architecture, knowledge, skills, tasks, methods, sensitive data, performance, human approval and versatile feedback... This survey extends findings from the robust community edited research papers BIG, BLOOM and HELM which open source, benchmark and analyze limits and challenges of large language models in terms of tasks complexity and strict evaluation on accuracy (e.g. fairness, robustness, toxicity, ...). It identifies the key elements used with Large Language Models (LLM) to solve complex questions or problems. Recent projects like ChatGPT and GALACTICA have allowed non-specialists to grasp the great potential as well as the equally strong limitations of language models in complex QA. Hybridizing these models with different components could allow to overcome these different limits and go much further. We discuss some challenges associated with complex QA, including domain adaptation, decomposition and efficient multi-step QA, long form QA, non-factoid QA, safety and multi-sensitivity data protection, multimodal search, hallucinations, QA explainability and truthfulness, time dimension. Therefore we review current solutions and promising strategies, using elements such as hybrid LLM architectures, human-in-the-loop reinforcement learning, prompting adaptation, neuro-symbolic and structured knowledge grounding, program synthesis, and others. We analyze existing solutions and provide an overview of the current research and trends in the area of complex QA.
Machine Learning
☆ Restoration-Degradation Beyond Linear Diffusions: A Non-Asymptotic Analysis For DDIM-Type Samplers
We develop a framework for non-asymptotic analysis of deterministic samplers used for diffusion generative modeling. Several recent works have analyzed stochastic samplers using tools like Girsanov's theorem and a chain rule variant of the interpolation argument. Unfortunately, these techniques give vacuous bounds when applied to deterministic samplers. We give a new operational interpretation for deterministic sampling by showing that one step along the probability flow ODE can be expressed as two steps: 1) a restoration step that runs gradient ascent on the conditional log-likelihood at some infinitesimally previous time, and 2) a degradation step that runs the forward process using noise pointing back towards the current iterate. This perspective allows us to extend denoising diffusion implicit models to general, non-linear forward processes. We then develop the first polynomial convergence bounds for these samplers under mild conditions on the data distribution.
comment: 29 pages
☆ Globally Optimal Training of Neural Networks with Threshold Activation Functions ICLR 2023
Threshold activation functions are highly preferable in neural networks due to their efficiency in hardware implementations. Moreover, their mode of operation is more interpretable and resembles that of biological neurons. However, traditional gradient based algorithms such as Gradient Descent cannot be used to train the parameters of neural networks with threshold activations since the activation function has zero gradient except at a single non-differentiable point. To this end, we study weight decay regularized training problems of deep neural networks with threshold activations. We first show that regularized deep threshold network training problems can be equivalently formulated as a standard convex optimization problem, which parallels the LASSO method, provided that the last hidden layer width exceeds a certain threshold. We also derive a simplified convex optimization formulation when the dataset can be shattered at a certain layer of the network. We corroborate our theoretical results with various numerical experiments.
comment: Accepted to ICLR 2023
☆ Learning Humanoid Locomotion with Transformers
We present a sim-to-real learning-based approach for real-world humanoid locomotion. Our controller is a causal Transformer trained by autoregressive prediction of future actions from the history of observations and actions. We hypothesize that the observation-action history contains useful information about the world that a powerful Transformer model can use to adapt its behavior in-context, without updating its weights. We do not use state estimation, dynamics models, trajectory optimization, reference trajectories, or pre-computed gait libraries. Our controller is trained with large-scale model-free reinforcement learning on an ensemble of randomized environments in simulation and deployed to the real world in a zero-shot fashion. We evaluate our approach in high-fidelity simulation and successfully deploy it to the real robot as well. To the best of our knowledge, this is the first demonstration of a fully learning-based method for real-world full-sized humanoid locomotion.
comment: Project page: https://humanoid-transformer.github.io
☆ SUREL+: Moving from Walks to Sets for Scalable Subgraph-based Graph Representation Learning
Subgraph-based graph representation learning (SGRL) has recently emerged as a powerful tool in many prediction tasks on graphs due to its advantages in model expressiveness and generalization ability. Most previous SGRL models face computational issues associated with the high cost of extracting subgraphs for each training or testing query. Recently, SUREL has been proposed as a new framework to accelerate SGRL, which samples random walks offline and joins these walks as subgraphs online for prediction. Due to the reusability of sampled walks across different queries, SUREL achieves state-of-the-art performance in both scalability and prediction accuracy. However, SUREL still suffers from high computational overhead caused by node redundancy in sampled walks. In this work, we propose a novel framework SUREL+ that upgrades SUREL by using node sets instead of walks to represent subgraphs. This set-based representation avoids node duplication by definition, but the sizes of node sets can be irregular. To address this issue, we design a dedicated sparse data structure to efficiently store and fast index node sets, and provide a specialized operator to join them in parallel batches. SUREL+ is modularized to support multiple types of set samplers, structural features, and neural encoders to complement the loss of structural information due to the reduction from walks to sets. Extensive experiments have been performed to validate SUREL+ in the prediction tasks of links, relation types, and higher-order patterns. SUREL+ achieves 3-11$\times$ speedups of SUREL while maintaining comparable or even better prediction performance; compared to other SGRL baselines, SUREL+ achieves $\sim$20$\times$ speedups and significantly improves the prediction accuracy.
☆ PaLM-E: An Embodied Multimodal Language Model
Large language models excel at a wide range of complex tasks. However, enabling general inference in the real world, e.g., for robotics problems, raises the challenge of grounding. We propose embodied language models to directly incorporate real-world continuous sensor modalities into language models and thereby establish the link between words and percepts. Input to our embodied language model are multi-modal sentences that interleave visual, continuous state estimation, and textual input encodings. We train these encodings end-to-end, in conjunction with a pre-trained large language model, for multiple embodied tasks including sequential robotic manipulation planning, visual question answering, and captioning. Our evaluations show that PaLM-E, a single large embodied multimodal model, can address a variety of embodied reasoning tasks, from a variety of observation modalities, on multiple embodiments, and further, exhibits positive transfer: the model benefits from diverse joint training across internet-scale language, vision, and visual-language domains. Our largest model, PaLM-E-562B with 562B parameters, in addition to being trained on robotics tasks, is a visual-language generalist with state-of-the-art performance on OK-VQA, and retains generalist language capabilities with increasing scale.
☆ MAESTRO: Open-Ended Environment Design for Multi-Agent Reinforcement Learning ICLR
Open-ended learning methods that automatically generate a curriculum of increasingly challenging tasks serve as a promising avenue toward generally capable reinforcement learning agents. Existing methods adapt curricula independently over either environment parameters (in single-agent settings) or co-player policies (in multi-agent settings). However, the strengths and weaknesses of co-players can manifest themselves differently depending on environmental features. It is thus crucial to consider the dependency between the environment and co-player when shaping a curriculum in multi-agent domains. In this work, we use this insight and extend Unsupervised Environment Design (UED) to multi-agent environments. We then introduce Multi-Agent Environment Design Strategist for Open-Ended Learning (MAESTRO), the first multi-agent UED approach for two-player zero-sum settings. MAESTRO efficiently produces adversarial, joint curricula over both environments and co-players and attains minimax-regret guarantees at Nash equilibrium. Our experiments show that MAESTRO outperforms a number of strong baselines on competitive two-player games, spanning discrete and continuous control settings.
comment: International Conference on Learning Representations (ICLR) 2023
☆ To Stay or Not to Stay in the Pre-train Basin: Insights on Ensembling in Transfer Learning
Transfer learning and ensembling are two popular techniques for improving the performance and robustness of neural networks. Due to the high cost of pre-training, ensembles of models fine-tuned from a single pre-trained checkpoint are often used in practice. Such models end up in the same basin of the loss landscape and thus have limited diversity. In this work, we study if it is possible to improve ensembles trained from a single pre-trained checkpoint by better exploring the pre-train basin or a close vicinity outside of it. We show that while exploration of the pre-train basin may be beneficial for the ensemble, leaving the basin results in losing the benefits of transfer learning and degradation of the ensemble quality.
comment: First two authors contributed equally
☆ ALMOST: Adversarial Learning to Mitigate Oracle-less ML Attacks via Synthesis Tuning
Oracle-less machine learning (ML) attacks have broken various logic locking schemes. Regular synthesis, which is tailored for area-power-delay optimization, yields netlists where key-gate localities are vulnerable to learning. Thus, we call for security-aware logic synthesis. We propose ALMOST, a framework for adversarial learning to mitigate oracle-less ML attacks via synthesis tuning. ALMOST uses a simulated-annealing-based synthesis recipe generator, employing adversarially trained models that can predict state-of-the-art attacks' accuracies over wide ranges of recipes and key-gate localities. Experiments on ISCAS benchmarks confirm the attacks' accuracies drops to around 50\% for ALMOST-synthesized circuits, all while not undermining design optimization.
comment: Accepted at Design Automation Conference (DAC 2023)
☆ Efficient Skill Acquisition for Complex Manipulation Tasks in Obstructed Environments
Data efficiency in robotic skill acquisition is crucial for operating robots in varied small-batch assembly settings. To operate in such environments, robots must have robust obstacle avoidance and versatile goal conditioning acquired from only a few simple demonstrations. Existing approaches, however, fall short of these requirements. Deep reinforcement learning (RL) enables a robot to learn complex manipulation tasks but is often limited to small task spaces in the real world due to sample inefficiency and safety concerns. Motion planning (MP) can generate collision-free paths in obstructed environments, but cannot solve complex manipulation tasks and requires goal states often specified by a user or object-specific pose estimator. In this work, we propose a system for efficient skill acquisition that leverages an object-centric generative model (OCGM) for versatile goal identification to specify a goal for MP combined with RL to solve complex manipulation tasks in obstructed environments. Specifically, OCGM enables one-shot target object identification and re-identification in new scenes, allowing MP to guide the robot to the target object while avoiding obstacles. This is combined with a skill transition network, which bridges the gap between terminal states of MP and feasible start states of a sample-efficient RL policy. The experiments demonstrate that our OCGM-based one-shot goal identification provides competitive accuracy to other baseline approaches and that our modular framework outperforms competitive baselines, including a state-of-the-art RL algorithm, by a significant margin for complex manipulation tasks in obstructed environments.
comment: 8 pages, 5 figures
☆ Leveraging Scene Embeddings for Gradient-Based Motion Planning in Latent Space
Motion planning framed as optimisation in structured latent spaces has recently emerged as competitive with traditional methods in terms of planning success while significantly outperforming them in terms of computational speed. However, the real-world applicability of recent work in this domain remains limited by the need to express obstacle information directly in state-space, involving simple geometric primitives. In this work we address this challenge by leveraging learned scene embeddings together with a generative model of the robot manipulator to drive the optimisation process. In addition, we introduce an approach for efficient collision checking which directly regularises the optimisation undertaken for planning. Using simulated as well as real-world experiments, we demonstrate that our approach, AMP-LS, is able to successfully plan in novel, complex scenes while outperforming traditional planning baselines in terms of computation speed by an order of magnitude. We show that the resulting system is fast enough to enable closed-loop planning in real-world dynamic scenes.
comment: Project website: https://amp-ls.github.io/
☆ Enhancing Activity Prediction Models in Drug Discovery with the Ability to Understand Human Language
Activity and property prediction models are the central workhorses in drug discovery and materials sciences, but currently they have to be trained or fine-tuned for new tasks. Without training or fine-tuning, scientific language models could be used for such low-data tasks through their announced zero- and few-shot capabilities. However, their predictive quality at activity prediction is lacking. In this work, we envision a novel type of activity prediction model that is able to adapt to new prediction tasks at inference time, via understanding textual information describing the task. To this end, we propose a new architecture with separate modules for chemical and natural language inputs, and a contrastive pre-training objective on data from large biochemical databases. In extensive experiments, we show that our method CLAMP yields improved predictive performance on few-shot learning benchmarks and zero-shot problems in drug discovery. We attribute the advances of our method to the modularized architecture and to our pre-training objective.
comment: 15 pages + 18 pages appendix
☆ Thompson Sampling for Linear Bandit Problems with Normal-Gamma Priors
We consider Thompson sampling for linear bandit problems with finitely many independent arms, where rewards are sampled from normal distributions that are linearly dependent on unknown parameter vectors and with unknown variance. Specifically, with a Bayesian formulation we consider multivariate normal-gamma priors to represent environment uncertainty for all involved parameters. We show that our chosen sampling prior is a conjugate prior to the reward model and derive a Bayesian regret bound for Thompson sampling under the condition that the 5/2-moment of the variance distribution exist.
comment: 27 pages, 2 figures
☆ Symbolic Synthesis of Neural Networks
Neural networks adapt very well to distributed and continuous representations, but struggle to learn and generalize from small amounts of data. Symbolic systems commonly achieve data efficient generalization by exploiting modularity to benefit from local and discrete features of a representation. These features allow symbolic programs to be improved one module at a time and to experience combinatorial growth in the values they can successfully process. However, it is difficult to design components that can be used to form symbolic abstractions and which are highly-overparametrized like neural networks, as the adjustment of parameters makes the semantics of modules unstable. I present Graph-based Symbolically Synthesized Neural Networks (G-SSNNs), a form of neural network whose topology and parameters are informed by the output of a symbolic program. I demonstrate that by developing symbolic abstractions at a population level, and applying gradient-based optimization to such neural models at an individual level, I can elicit reliable patterns of improved generalization with small quantities of data known to contain local and discrete features. The paradigm embodied by G-SSNNs offers a route towards the communal development of compact and composable abstractions which can be flexibly repurposed for a variety of tasks and high-dimensional media. In future work, I hope to pursue these benefits by exploring more ambitious G-SSNN designs based on more complex classes of symbolic programs. The code and data associated with the reported results are publicly available at https://github.com/shlomenu/symbolically_synthesized_networks .
comment: 8 pages, 1 figure
☆ Lower Bounds for $γ$-Regret via the Decision-Estimation Coefficient
In this note, we give a new lower bound for the $\gamma$-regret in bandit problems, the regret which arises when comparing against a benchmark that is $\gamma$ times the optimal solution, i.e., $\mathsf{Reg}_{\gamma}(T) = \sum_{t = 1}^T \gamma \max_{\pi} f(\pi) - f(\pi_t)$. The $\gamma$-regret arises in structured bandit problems where finding an exact optimum of $f$ is intractable. Our lower bound is given in terms of a modification of the constrained Decision-Estimation Coefficient (DEC) of~\citet{foster2023tight} (and closely related to the original offset DEC of \citet{foster2021statistical}), which we term the $\gamma$-DEC. When restricted to the traditional regret setting where $\gamma = 1$, our result removes the logarithmic factors in the lower bound of \citet{foster2023tight}.
☆ Keep It Simple: CNN Model Complexity Studies for Interference Classification Tasks
The growing number of devices using the wireless spectrum makes it important to find ways to minimize interference and optimize the use of the spectrum. Deep learning models, such as convolutional neural networks (CNNs), have been widely utilized to identify, classify, or mitigate interference due to their ability to learn from the data directly. However, there have been limited research on the complexity of such deep learning models. The major focus of deep learning-based wireless classification literature has been on improving classification accuracy, often at the expense of model complexity. This may not be practical for many wireless devices, such as, internet of things (IoT) devices, which usually have very limited computational resources and cannot handle very complex models. Thus, it becomes important to account for model complexity when designing deep learning-based models for interference classification. To address this, we conduct an analysis of CNN based wireless classification that explores the trade-off amongst dataset size, CNN model complexity, and classification accuracy under various levels of classification difficulty: namely, interference classification, heterogeneous transmitter classification, and homogeneous transmitter classification. Our study, based on three wireless datasets, shows that a simpler CNN model with fewer parameters can perform just as well as a more complex model, providing important insights into the use of CNNs in computationally constrained applications.
comment: 6 pages, 7 figures, 3 tables
☆ Time series anomaly detection with sequence reconstruction based state-space model
Recent advances in digitization has led to availability of multivariate time series data in various domains, in order to monitor operations in real time. Identifying abnormal data pattern and detect potential failures in these scenarios are important yet rather difficult tasks. We propose a novel unsupervised anomaly detection method for time series data. Our approach uses sequence encoder and decoder to represent the mapping between time series and hidden state, and learns bidirectional dynamics simultaneously by leveraging backward and forward temporal information in the training process. We further regularize the state space to place constraints on states of normal samples, and use Mahalanobis distance to evaluate abnormality level. Results on synthetic and real-world datasets show the superiority of the proposed method.
☆ CleanCLIP: Mitigating Data Poisoning Attacks in Multimodal Contrastive Learning
Multimodal contrastive pretraining has been utilized to train multimodal representation models, like CLIP, on vast amounts of paired image-text data. However, previous studies have highlighted the susceptibility of such models to backdoor attacks. Specifically, when training on backdoored examples, CLIP learns spurious correlations between the embedded backdoor trigger and the target label, aligning their representations in the joint embedding space. With injecting only a few poisoned examples e.g., 75 examples in the 3M pretraining data, the model's behavior can be significantly manipulated, thus making it hard to detect or unlearn such correlations. To address this issue, we propose CleanCLIP, a finetuning framework that weakens the learned spurious associations introduced by backdoor attacks by re-aligning the representations for individual modalities independently. CleanCLIP can be employed for both unsupervised finetuning on paired image-text data and for supervised finetuning on labeled image data. We demonstrate that unsupervised finetuning with a combination of multimodal contrastive and unimodal self-supervised objectives for individual modalities can significantly reduce the impact of the backdoor attack. Additionally, supervised finetuning on task-specific labeled data of the individual modality, such as image data, removes the backdoor trigger from the CLIP vision encoder. Empirically, we show that CleanCLIP maintains model performance on benign examples while mitigating the impact of a range of backdoor attacks on multimodal contrastive learning.
comment: 20 pages, 7 figures, 8 tables
☆ Learning to Backdoor Federated Learning
In a federated learning (FL) system, malicious participants can easily embed backdoors into the aggregated model while maintaining the model's performance on the main task. To this end, various defenses, including training stage aggregation-based defenses and post-training mitigation defenses, have been proposed recently. While these defenses obtain reasonable performance against existing backdoor attacks, which are mainly heuristics based, we show that they are insufficient in the face of more advanced attacks. In particular, we propose a general reinforcement learning-based backdoor attack framework where the attacker first trains a (non-myopic) attack policy using a simulator built upon its local data and common knowledge on the FL system, which is then applied during actual FL training. Our attack framework is both adaptive and flexible and achieves strong attack performance and durability even under state-of-the-art defenses.
☆ Weight Perturbation Can Help Fairness under Distribution Shift
Fairness in machine learning has attracted increasing attention in recent years. The fairness methods improving algorithmic fairness for in-distribution data may not perform well under distribution shift. In this paper, we first theoretically demonstrate the inherent connection between distribution shift, data perturbation, and weight perturbation. Subsequently, we analyze the sufficient conditions to guarantee fairness (i.e., low demographic parity) for the target dataset, including fairness for the source dataset, and low prediction difference between the source and target dataset for each sensitive attribute group. Motivated by these sufficient conditions, we propose robust fairness regularization (RFR) by considering the worst case within the weight perturbation ball for each sensitive attribute group. In this way, the maximization problem can be simplified as two forward and two backward propagations for each update of model parameters. We evaluate the effectiveness of our proposed RFR algorithm on synthetic and real distribution shifts across various datasets. Experimental results demonstrate that RFR achieves better fairness-accuracy trade-off performance compared with several baselines.
☆ HiGeN: Hierarchical Multi-Resolution Graph Generative Networks
In real world domains, most graphs naturally exhibit a hierarchical structure. However, data-driven graph generation is yet to effectively capture such structures. To address this, we propose a novel approach that recursively generates community structures at multiple resolutions, with the generated structures conforming to training data distribution at each level of the hierarchy. The graphs generation is designed as a sequence of coarse-to-fine generative models allowing for parallel generation of all sub-structures, resulting in a high degree of scalability. Furthermore, we model the output distribution of edges with a more expressive multinomial distribution and derive a recursive factorization for this distribution, making it a suitable choice for graph generative models. This allows for the generation of graphs with integer-valued edge weights. Our method achieves state-of-the-art performance in both accuracy and efficiency on multiple datasets.
☆ The Wasserstein Believer: Learning Belief Updates for Partially Observable Environments through Reliable Latent Space Models
Partially Observable Markov Decision Processes (POMDPs) are useful tools to model environments where the full state cannot be perceived by an agent. As such the agent needs to reason taking into account the past observations and actions. However, simply remembering the full history is generally intractable due to the exponential growth in the history space. Keeping a probability distribution that models the belief over what the true state is can be used as a sufficient statistic of the history, but its computation requires access to the model of the environment and is also intractable. Current state-of-the-art algorithms use Recurrent Neural Networks (RNNs) to compress the observation-action history aiming to learn a sufficient statistic, but they lack guarantees of success and can lead to suboptimal policies. To overcome this, we propose the Wasserstein-Belief-Updater (WBU), an RL algorithm that learns a latent model of the POMDP and an approximation of the belief update. Our approach comes with theoretical guarantees on the quality of our approximation ensuring that our outputted beliefs allow for learning the optimal value function.
☆ Faithfulness-Aware Decoding Strategies for Abstractive Summarization EACL 2023
Despite significant progress in understanding and improving faithfulness in abstractive summarization, the question of how decoding strategies affect faithfulness is less studied. We present a systematic study of the effect of generation techniques such as beam search and nucleus sampling on faithfulness in abstractive summarization. We find a consistent trend where beam search with large beam sizes produces the most faithful summaries while nucleus sampling generates the least faithful ones. We propose two faithfulness-aware generation methods to further improve faithfulness over current generation techniques: (1) ranking candidates generated by beam search using automatic faithfulness metrics and (2) incorporating lookahead heuristics that produce a faithfulness score on the future summary. We show that both generation methods significantly improve faithfulness across two datasets as evaluated by four automatic faithfulness metrics and human evaluation. To reduce computational cost, we demonstrate a simple distillation approach that allows the model to generate faithful summaries with just greedy decoding. Our code is publicly available at https://github.com/amazon-science/faithful-summarization-generation
comment: EACL 2023 (17 pages)
☆ Deep symbolic regression for physics guided by units constraints: toward the automated discovery of physical laws
Symbolic Regression is the study of algorithms that automate the search for analytic expressions that fit data. While recent advances in deep learning have generated renewed interest in such approaches, efforts have not been focused on physics, where we have important additional constraints due to the units associated with our data. Here we present $\Phi$-SO, a Physical Symbolic Optimization framework for recovering analytical symbolic expressions from physics data using deep reinforcement learning techniques by learning units constraints. Our system is built, from the ground up, to propose solutions where the physical units are consistent by construction. This is useful not only in eliminating physically impossible solutions, but because it restricts enormously the freedom of the equation generator, thus vastly improving performance. The algorithm can be used to fit noiseless data, which can be useful for instance when attempting to derive an analytical property of a physical model, and it can also be used to obtain analytical approximations to noisy data. We showcase our machinery on a panel of examples from astrophysics.
comment: 16 pages, 6 figures, 2 tables. Submitted to ApJ
☆ Accelerated Rates between Stochastic and Adversarial Online Convex Optimization
Stochastic and adversarial data are two widely studied settings in online learning. But many optimization tasks are neither i.i.d. nor fully adversarial, which makes it of fundamental interest to get a better theoretical understanding of the world between these extremes. In this work we establish novel regret bounds for online convex optimization in a setting that interpolates between stochastic i.i.d. and fully adversarial losses. By exploiting smoothness of the expected losses, these bounds replace a dependence on the maximum gradient length by the variance of the gradients, which was previously known only for linear losses. In addition, they weaken the i.i.d. assumption by allowing, for example, adversarially poisoned rounds, which were previously considered in the related expert and bandit settings. In the fully i.i.d. case, our regret bounds match the rates one would expect from results in stochastic acceleration, and we also recover the optimal stochastically accelerated rates via online-to-batch conversion. In the fully adversarial case our bounds gracefully deteriorate to match the minimax regret. We further provide lower bounds showing that our regret upper bounds are tight for all intermediate regimes in terms of the stochastic variance and the adversarial variation of the loss gradients.
comment: Extended version of 'Between Stochastic and Adversarial Online Convex Optimization: Improved Regret Bounds via Smoothness' by the same authors. arXiv admin note: text overlap with arXiv:2202.07554
☆ An Online Algorithm for Chance Constrained Resource Allocation ICASSP 2023
This paper studies the online stochastic resource allocation problem (RAP) with chance constraints. The online RAP is a 0-1 integer linear programming problem where the resource consumption coefficients are revealed column by column along with the corresponding revenue coefficients. When a column is revealed, the corresponding decision variables are determined instantaneously without future information. Moreover, in online applications, the resource consumption coefficients are often obtained by prediction. To model their uncertainties, we take the chance constraints into the consideration. To the best of our knowledge, this is the first time chance constraints are introduced in the online RAP problem. Assuming that the uncertain variables have known Gaussian distributions, the stochastic RAP can be transformed into a deterministic but nonlinear problem with integer second-order cone constraints. Next, we linearize this nonlinear problem and analyze the performance of vanilla online primal-dual algorithm for solving the linearized stochastic RAP. Under mild technical assumptions, the optimality gap and constraint violation are both on the order of $\sqrt{n}$. Then, to further improve the performance of the algorithm, several modified online primal-dual algorithms with heuristic corrections are proposed. Finally, extensive numerical experiments on both synthetic and real data demonstrate the applicability and effectiveness of our methods.
comment: 5 pages, 5 figures. Accepted to ICASSP 2023. arXiv admin note: substantial text overlap with arXiv:2203.16818
☆ Evaluating the Fairness of Deep Learning Uncertainty Estimates in Medical Image Analysis
Although deep learning (DL) models have shown great success in many medical image analysis tasks, deployment of the resulting models into real clinical contexts requires: (1) that they exhibit robustness and fairness across different sub-populations, and (2) that the confidence in DL model predictions be accurately expressed in the form of uncertainties. Unfortunately, recent studies have indeed shown significant biases in DL models across demographic subgroups (e.g., race, sex, age) in the context of medical image analysis, indicating a lack of fairness in the models. Although several methods have been proposed in the ML literature to mitigate a lack of fairness in DL models, they focus entirely on the absolute performance between groups without considering their effect on uncertainty estimation. In this work, we present the first exploration of the effect of popular fairness models on overcoming biases across subgroups in medical image analysis in terms of bottom-line performance, and their effects on uncertainty quantification. We perform extensive experiments on three different clinically relevant tasks: (i) skin lesion classification, (ii) brain tumour segmentation, and (iii) Alzheimer's disease clinical score regression. Our results indicate that popular ML methods, such as data-balancing and distributionally robust optimization, succeed in mitigating fairness issues in terms of the model performances for some of the tasks. However, this can come at the cost of poor uncertainty estimates associated with the model predictions. This tradeoff must be mitigated if fairness models are to be adopted in medical image analysis.
comment: Paper accepted at MIDL 2023
☆ Convergence Rates for Non-Log-Concave Sampling and Log-Partition Estimation
Sampling from Gibbs distributions $p(x) \propto \exp(-V(x)/\varepsilon)$ and computing their log-partition function are fundamental tasks in statistics, machine learning, and statistical physics. However, while efficient algorithms are known for convex potentials $V$, the situation is much more difficult in the non-convex case, where algorithms necessarily suffer from the curse of dimensionality in the worst case. For optimization, which can be seen as a low-temperature limit of sampling, it is known that smooth functions $V$ allow faster convergence rates. Specifically, for $m$-times differentiable functions in $d$ dimensions, the optimal rate for algorithms with $n$ function evaluations is known to be $O(n^{-m/d})$, where the constant can potentially depend on $m, d$ and the function to be optimized. Hence, the curse of dimensionality can be alleviated for smooth functions at least in terms of the convergence rate. Recently, it has been shown that similarly fast rates can also be achieved with polynomial runtime $O(n^{3.5})$, where the exponent $3.5$ is independent of $m$ or $d$. Hence, it is natural to ask whether similar rates for sampling and log-partition computation are possible, and whether they can be realized in polynomial time with an exponent independent of $m$ and $d$. We show that the optimal rates for sampling and log-partition computation are sometimes equal and sometimes faster than for optimization. We then analyze various polynomial-time sampling algorithms, including an extension of a recent promising optimization approach, and find that they sometimes exhibit interesting behavior but no near-optimal rates. Our results also give further insights on the relation between sampling, log-partition, and optimization problems.
comment: Plots can be reproduced using the code at https://github.com/dholzmueller/sampling_experiments
☆ Parallel Hybrid Networks: an interplay between quantum and classical neural networks
Quantum neural networks represent a new machine learning paradigm that has recently attracted much attention due to its potential promise. Under certain conditions, these models approximate the distribution of their dataset with a truncated Fourier series. The trigonometric nature of this fit could result in angle-embedded quantum neural networks struggling to fit the non-harmonic features in a given dataset. Moreover, the interpretability of neural networks remains a challenge. In this work, we introduce a new, interpretable class of hybrid quantum neural networks that pass the inputs of the dataset in parallel to 1) a classical multi-layered perceptron and 2) a variational quantum circuit, and then the outputs of the two are linearly combined. We observe that the quantum neural network creates a smooth sinusoidal foundation base on the training set, and then the classical perceptrons fill the non-harmonic gaps in the landscape. We demonstrate this claim on two synthetic datasets sampled from periodic distributions with added protrusions as noise. The training results indicate that the parallel hybrid network architecture could improve the solution optimality on periodic datasets with additional noise.
comment: 8 pages, 6 figures
☆ Safe Reinforcement Learning via Probabilistic Logic Shields
Safe Reinforcement learning (Safe RL) aims at learning optimal policies while staying safe. A popular solution to Safe RL is shielding, which uses a logical safety specification to prevent an RL agent from taking unsafe actions. However, traditional shielding techniques are difficult to integrate with continuous, end-to-end deep RL methods. To this end, we introduce Probabilistic Logic Policy Gradient (PLPG). PLPG is a model-based Safe RL technique that uses probabilistic logic programming to model logical safety constraints as differentiable functions. Therefore, PLPG can be seamlessly applied to any policy gradient algorithm while still providing the same convergence guarantees. In our experiments, we show that PLPG learns safer and more rewarding policies compared to other state-of-the-art shielding techniques.
☆ Boosting Differentiable Causal Discovery via Adaptive Sample Reweighting ICLR 2023
Under stringent model type and variable distribution assumptions, differentiable score-based causal discovery methods learn a directed acyclic graph (DAG) from observational data by evaluating candidate graphs over an average score function. Despite great success in low-dimensional linear systems, it has been observed that these approaches overly exploit easier-to-fit samples, thus inevitably learning spurious edges. Worse still, inherent mostly in these methods the common homogeneity assumption can be easily violated, due to the widespread existence of heterogeneous data in the real world, resulting in performance vulnerability when noise distributions vary. We propose a simple yet effective model-agnostic framework to boost causal discovery performance by dynamically learning the adaptive weights for the Reweighted Score function, ReScore for short, where the weights tailor quantitatively to the importance degree of each sample. Intuitively, we leverage the bilevel optimization scheme to \wx{alternately train a standard DAG learner and reweight samples -- that is, upweight the samples the learner fails to fit and downweight the samples that the learner easily extracts the spurious information from. Extensive experiments on both synthetic and real-world datasets are carried out to validate the effectiveness of ReScore. We observe consistent and significant boosts in structure learning performance. Furthermore, we visualize that ReScore concurrently mitigates the influence of spurious edges and generalizes to heterogeneous data. Finally, we perform the theoretical analysis to guarantee the structure identifiability and the weight adaptive properties of ReScore in linear systems. Our codes are available at https://github.com/anzhang314/ReScore.
comment: In proceedings of ICLR 2023
☆ MetaPhysiCa: OOD Robustness in Physics-informed Machine Learning
A fundamental challenge in physics-informed machine learning (PIML) is the design of robust PIML methods for out-of-distribution (OOD) forecasting tasks. These OOD tasks require learning-to-learn from observations of the same (ODE) dynamical system with different unknown ODE parameters, and demand accurate forecasts even under out-of-support initial conditions and out-of-support ODE parameters. In this work we propose a solution for such tasks, which we define as a meta-learning procedure for causal structure discovery (including invariant risk minimization). Using three different OOD tasks, we empirically observe that the proposed approach significantly outperforms existing state-of-the-art PIML and deep learning methods.
☆ A Unified Algebraic Perspective on Lipschitz Neural Networks ICLR 2023
Important research efforts have focused on the design and training of neural networks with a controlled Lipschitz constant. The goal is to increase and sometimes guarantee the robustness against adversarial attacks. Recent promising techniques draw inspirations from different backgrounds to design 1-Lipschitz neural networks, just to name a few: convex potential layers derive from the discretization of continuous dynamical systems, Almost-Orthogonal-Layer proposes a tailored method for matrix rescaling. However, it is today important to consider the recent and promising contributions in the field under a common theoretical lens to better design new and improved layers. This paper introduces a novel algebraic perspective unifying various types of 1-Lipschitz neural networks, including the ones previously mentioned, along with methods based on orthogonality and spectral methods. Interestingly, we show that many existing techniques can be derived and generalized via finding analytical solutions of a common semidefinite programming (SDP) condition. We also prove that AOL biases the scaled weight to the ones which are close to the set of orthogonal matrices in a certain mathematical manner. Moreover, our algebraic condition, combined with the Gershgorin circle theorem, readily leads to new and diverse parameterizations for 1-Lipschitz network layers. Our approach, called SDP-based Lipschitz Layers (SLL), allows us to design non-trivial yet efficient generalization of convex potential layers. Finally, the comprehensive set of experiments on image classification shows that SLLs outperform previous approaches on certified robust accuracy. Code is available at https://github.com/araujoalexandre/Lipschitz-SLL-Networks.
comment: ICLR 2023. Spotlight paper
☆ Data-Driven Control with Inherent Lyapunov Stability
Recent advances in learning-based control leverage deep function approximators, such as neural networks, to model the evolution of controlled dynamical systems over time. However, the problem of learning a dynamics model and a stabilizing controller persists, since the synthesis of a stabilizing feedback law for known nonlinear systems is a difficult task, let alone for complex parametric representations that must be fit to data. To this end, we propose a method for jointly learning parametric representations of a nonlinear dynamics model and a stabilizing controller from data. To do this, our approach simultaneously learns a parametric Lyapunov function which intrinsically constrains the dynamics model to be stabilizable by the learned controller. In addition to the stabilizability of the learned dynamics guaranteed by our novel construction, we show that the learned controller stabilizes the true dynamics under certain assumptions on the fidelity of the learned dynamics. Finally, we demonstrate the efficacy of our method on a variety of simulated nonlinear dynamical systems.
☆ SC-Block: Supervised Contrastive Blocking within Entity Resolution Pipelines
The goal of entity resolution is to identify records in multiple datasets that represent the same real-world entity. However, comparing all records across datasets can be computationally intensive, leading to long runtimes. To reduce these runtimes, entity resolution pipelines are constructed of two parts: a blocker that applies a computationally cheap method to select candidate record pairs, and a matcher that afterwards identifies matching pairs from this set using more expensive methods. This paper presents SC-Block, a blocking method that utilizes supervised contrastive learning for positioning records in the embedding space, and nearest neighbour search for candidate set building. We benchmark SC-Block against eight state-of-the-art blocking methods. In order to relate the training time of SC-Block to the reduction of the overall runtime of the entity resolution pipeline, we combine SC-Block with four matching methods into complete pipelines. For measuring the overall runtime, we determine candidate sets with 98% pair completeness and pass them to the matcher. The results show that SC-Block is able to create smaller candidate sets and pipelines with SC-Block execute 1.5 to 2 times faster compared to pipelines with other blockers, without sacrificing F1 score. Blockers are often evaluated using relatively small datasets which might lead to runtime effects resulting from a large vocabulary size being overlooked. In order to measure runtimes in a more challenging setting, we introduce a new benchmark dataset that requires large numbers of product offers to be blocked. On this large-scale benchmark dataset, pipelines utilizing SC-Block and the best-performing matcher execute 8 times faster than pipelines utilizing another blocker with the same matcher reducing the runtime from 2.5 hours to 18 minutes, clearly compensating for the 5 minutes required for training SC-Block.
☆ Benchmark of Data Preprocessing Methods for Imbalanced Classification
Severe class imbalance is one of the main conditions that make machine learning in cybersecurity difficult. A variety of dataset preprocessing methods have been introduced over the years. These methods modify the training dataset by oversampling, undersampling or a combination of both to improve the predictive performance of classifiers trained on this dataset. Although these methods are used in cybersecurity occasionally, a comprehensive, unbiased benchmark comparing their performance over a variety of cybersecurity problems is missing. This paper presents a benchmark of 16 preprocessing methods on six cybersecurity datasets together with 17 public imbalanced datasets from other domains. We test the methods under multiple hyperparameter configurations and use an AutoML system to train classifiers on the preprocessed datasets, which reduces potential bias from specific hyperparameter or classifier choices. Special consideration is also given to evaluating the methods using appropriate performance measures that are good proxies for practical performance in real-world cybersecurity systems. The main findings of our study are: 1) Most of the time, a data preprocessing method that improves classification performance exists. 2) Baseline approach of doing nothing outperformed a large portion of methods in the benchmark. 3) Oversampling methods generally outperform undersampling methods. 4) The most significant performance gains are brought by the standard SMOTE algorithm and more complicated methods provide mainly incremental improvements at the cost of often worse computational performance.
comment: Published at 2022 IEEE International Conference on Big Data (Big Data) 2022. Extended version with full results in appendix
☆ Environment Invariant Linear Least Squares
This paper considers a multiple environments linear regression model in which data from multiple experimental settings are collected. The joint distribution of the response variable and covariate may vary across different environments, yet the conditional expectation of $y$ given the unknown set of important variables are invariant across environments. Such a statistical model is related to the problem of endogeneity, causal inference, and transfer learning. The motivation behind it is illustrated by how the goals of prediction and attribution are inherent in estimating the true parameter and the important variable set. We construct a novel {\it environment invariant linear least squares (EILLS)} objective function, a multiple-environment version of linear least squares that leverages the above conditional expectation invariance structure and heterogeneity among different environments to determine the true parameter. Our proposed method is applicable without any additional structural knowledge and can identify the true parameter under a near-minimal identification condition. We establish non-asymptotic $\ell_2$ error bounds on the estimation error for the EILLS estimator in the presence of spurious variables. Moreover, we further show that the EILLS estimator is able to eliminate all endogenous variables and the $\ell_0$ penalized EILLS estimator can achieve variable selection consistency in high-dimensional regimes. These non-asymptotic results demonstrate the sample efficiency of the EILLS estimator and its capability to circumvent the curse of endogeneity in an algorithmic manner without any prior structural knowledge.
comment: 65 pages, 4 figures
☆ On Regression in Extreme Regions
In the classic regression problem, the value of a real-valued random variable $Y$ is to be predicted based on the observation of a random vector $X$, taking its values in $\mathbb{R}^d$ with $d\geq 1$ say. The statistical learning problem consists in building a predictive function $\hat{f}:\mathbb{R}^d\to \mathbb{R}$ based on independent copies of the pair $(X,Y)$ so that $Y$ is approximated by $\hat{f}(X)$ with minimum error in the mean-squared sense. Motivated by various applications, ranging from environmental sciences to finance or insurance, special attention is paid here to the case of extreme (i.e. very large) observations $X$. Because of their rarity, they contribute in a negligible manner to the (empirical) error and the predictive performance of empirical quadratic risk minimizers can be consequently very poor in extreme regions. In this paper, we develop a general framework for regression in the extremes. It is assumed that $X$'s conditional distribution given $Y$ belongs to a non parametric class of heavy-tailed probability distributions. It is then shown that an asymptotic notion of risk can be tailored to summarize appropriately predictive performance in extreme regions of the input space. It is also proved that minimization of an empirical and non asymptotic version of this 'extreme risk', based on a fraction of the largest observations solely, yields regression functions with good generalization capacity. In addition, numerical results providing strong empirical evidence of the relevance of the approach proposed are displayed.
comment: 10 pages (main paper), 10 pages (appendix)
☆ A neural network based model for multi-dimensional nonlinear Hawkes processes
This paper introduces the Neural Network for Nonlinear Hawkes processes (NNNH), a non-parametric method based on neural networks to fit nonlinear Hawkes processes. Our method is suitable for analyzing large datasets in which events exhibit both mutually-exciting and inhibitive patterns. The NNNH approach models the individual kernels and the base intensity of the nonlinear Hawkes process using feed forward neural networks and jointly calibrates the parameters of the networks by maximizing the log-likelihood function. We utilize Stochastic Gradient Descent to search for the optimal parameters and propose an unbiased estimator for the gradient, as well as an efficient computation method. We demonstrate the flexibility and accuracy of our method through numerical experiments on both simulated and real-world data, and compare it with state-of-the-art methods. Our results highlight the effectiveness of the NNNH method in accurately capturing the complexities of nonlinear Hawkes processes.
☆ Reinforcement Learning Based Self-play and State Stacking Techniques for Noisy Air Combat Environment
Reinforcement learning (RL) has recently proven itself as a powerful instrument for solving complex problems and even surpassed human performance in several challenging applications. This signifies that RL algorithms can be used in the autonomous air combat problem, which has been studied for many years. The complexity of air combat arises from aggressive close-range maneuvers and agile enemy behaviors. In addition to these complexities, there may be uncertainties in real-life scenarios due to sensor errors, which prevent estimation of the actual position of the enemy. In this case, autonomous aircraft should be successful even in the noisy environments. In this study, we developed an air combat simulation, which provides noisy observations to the agents, therefore, make the air combat problem even more challenging. Thus, we present a state stacking method for noisy RL environments as a noise reduction technique. In our extensive set of experiments, the proposed method significantly outperforms the baseline algorithms in terms of the winning ratio, where the performance improvement is even more pronounced in the high noise levels. In addition, we incorporate a self-play scheme to our training process by periodically updating the enemy with a frozen copy of the training agent. By this way, the training agent performs air combat simulations to an enemy with smarter strategies, which improves the performance and robustness of the agents. In our simulations, we demonstrate that the self-play scheme provides important performance gains compared to the classical RL training.
comment: 10 pages, 4 figures
☆ Masked Images Are Counterfactual Samples for Robust Fine-tuning CVPR 2023
Deep learning models are challenged by the distribution shift between the training data and test data. Recently, the large models pre-trained on diverse data demonstrate unprecedented robustness to various distribution shifts. However, fine-tuning on these models can lead to a trade-off between in-distribution (ID) performance and out-of-distribution (OOD) robustness. Existing methods for tackling this trade-off do not explicitly address the OOD robustness problem. In this paper, based on causal analysis on the aforementioned problems, we propose a novel fine-tuning method, which use masked images as counterfactual samples that help improving the robustness of the fine-tuning model. Specifically, we mask either the semantics-related or semantics-unrelated patches of the images based on class activation map to break the spurious correlation, and refill the masked patches with patches from other images. The resulting counterfactual samples are used in feature-based distillation with the pre-trained model. Extensive experiments verify that regularizing the fine-tuning with the proposed masked images can achieve a better trade-off between ID and OOD, surpassing previous methods on the OOD performance. Our code will be publicly available.
comment: Accepted by CVPR 2023
☆ Convolutional Neural Networks as 2-D systems
This paper introduces a novel representation of convolutional Neural Networks (CNNs) in terms of 2-D dynamical systems. To this end, the usual description of convolutional layers with convolution kernels, i.e., the impulse responses of linear filters, is realized in state space as a linear time-invariant 2-D system. The overall convolutional Neural Network composed of convolutional layers and nonlinear activation functions is then viewed as a 2-D version of a Lur'e system, i.e., a linear dynamical system interconnected with static nonlinear components. One benefit of this 2-D Lur'e system perspective on CNNs is that we can use robust control theory much more efficiently for Lipschitz constant estimation than previously possible.
☆ Deep Clustering with a Constraint for Topological Invariance based on Symmetric InfoNCE
We consider the scenario of deep clustering, in which the available prior knowledge is limited. In this scenario, few existing state-of-the-art deep clustering methods can perform well for both non-complex topology and complex topology datasets. To address the problem, we propose a constraint utilizing symmetric InfoNCE, which helps an objective of deep clustering method in the scenario train the model so as to be efficient for not only non-complex topology but also complex topology datasets. Additionally, we provide several theoretical explanations of the reason why the constraint can enhances performance of deep clustering methods. To confirm the effectiveness of the proposed constraint, we introduce a deep clustering method named MIST, which is a combination of an existing deep clustering method and our constraint. Our numerical experiments via MIST demonstrate that the constraint is effective. In addition, MIST outperforms other state-of-the-art deep clustering methods for most of the commonly used ten benchmark datasets.
comment: 48 pages, 6 figures
☆ Critical Points and Convergence Analysis of Generative Deep Linear Networks Trained with Bures-Wasserstein Loss
We consider a deep matrix factorization model of covariance matrices trained with the Bures-Wasserstein distance. While recent works have made important advances in the study of the optimization problem for overparametrized low-rank matrix approximation, much emphasis has been placed on discriminative settings and the square loss. In contrast, our model considers another interesting type of loss and connects with the generative setting. We characterize the critical points and minimizers of the Bures-Wasserstein distance over the space of rank-bounded matrices. For low-rank matrices the Hessian of this loss can theoretically blow up, which creates challenges to analyze convergence of optimizaton methods. We establish convergence results for gradient flow using a smooth perturbative version of the loss and convergence results for finite step size gradient descent under certain assumptions on the initial weights.
comment: 35 pages, 1 figure
☆ Guiding Energy-based Models via Contrastive Latent Variables ICLR 2023
An energy-based model (EBM) is a popular generative framework that offers both explicit density and architectural flexibility, but training them is difficult since it is often unstable and time-consuming. In recent years, various training techniques have been developed, e.g., better divergence measures or stabilization in MCMC sampling, but there often exists a large gap between EBMs and other generative frameworks like GANs in terms of generation quality. In this paper, we propose a novel and effective framework for improving EBMs via contrastive representation learning (CRL). To be specific, we consider representations learned by contrastive methods as the true underlying latent variable. This contrastive latent variable could guide EBMs to understand the data structure better, so it can improve and accelerate EBM training significantly. To enable the joint training of EBM and CRL, we also design a new class of latent-variable EBMs for learning the joint density of data and the contrastive latent variable. Our experimental results demonstrate that our scheme achieves lower FID scores, compared to prior-art EBM methods (e.g., additionally using variational autoencoders or diffusion techniques), even with significantly faster and more memory-efficient training. We also show conditional and compositional generation abilities of our latent-variable EBMs as their additional benefits, even without explicit conditional training. The code is available at https://github.com/hankook/CLEL.
comment: Accepted to ICLR 2023 (Spotlight). The code is available at https://github.com/hankook/CLEL
☆ Scaling strategies for on-device low-complexity source separation with Conv-Tasnet
Recently, several very effective neural approaches for single-channel speech separation have been presented in the literature. However, due to the size and complexity of these models, their use on low-resource devices, e.g. for hearing aids, and earphones, is still a challenge and established solutions are not available yet. Although approaches based on either pruning or compressing neural models have been proposed, the design of a model architecture suitable for a certain application domain often requires heuristic procedures not easily portable to different low-resource platforms. Given the modular nature of the well-known Conv-Tasnet speech separation architecture, in this paper we consider three parameters that directly control the overall size of the model, namely: the number of residual blocks, the number of repetitions of the separation blocks and the number of channels in the depth-wise convolutions, and experimentally evaluate how they affect the speech separation performance. In particular, experiments carried out on the Libri2Mix show that the number of dilated 1D-Conv blocks is the most critical parameter and that the usage of extra-dilation in the residual blocks allows reducing the performance drop.
☆ HiCLIP: Contrastive Language-Image Pretraining with Hierarchy-aware Attention ICLR 2023
The success of large-scale contrastive vision-language pretraining (CLIP) has benefited both visual recognition and multimodal content understanding. The concise design brings CLIP the advantage in inference efficiency against other vision-language models with heavier cross-attention fusion layers, making it a popular choice for a wide spectrum of downstream tasks. However, CLIP does not explicitly capture the hierarchical nature of high-level and fine-grained semantics conveyed in images and texts, which is arguably critical to vision-language understanding and reasoning. To this end, we equip both the visual and language branches in CLIP with hierarchy-aware attentions, namely Hierarchy-aware CLIP (HiCLIP), to progressively discover semantic hierarchies layer-by-layer from both images and texts in an unsupervised manner. As a result, such hierarchical aggregation significantly improves the cross-modal alignment. To demonstrate the advantages of HiCLIP, we conduct qualitative analysis on its unsupervised hierarchy induction during inference, as well as extensive quantitative experiments on both visual recognition and vision-language downstream tasks.
comment: Accepted at ICLR 2023
☆ Searching for Effective Neural Network Architectures for Heart Murmur Detection from Phonocardiogram
Aim: The George B. Moody PhysioNet Challenge 2022 raised problems of heart murmur detection and related abnormal cardiac function identification from phonocardiograms (PCGs). This work describes the novel approaches developed by our team, Revenger, to solve these problems. Methods: PCGs were resampled to 1000 Hz, then filtered with a Butterworth band-pass filter of order 3, cutoff frequencies 25 - 400 Hz, and z-score normalized. We used the multi-task learning (MTL) method via hard parameter sharing to train one neural network (NN) model for all the Challenge tasks. We performed neural architecture searching among a set of network backbones, including multi-branch convolutional neural networks (CNNs), SE-ResNets, TResNets, simplified wav2vec2, etc. Based on a stratified splitting of the subjects, 20% of the public data was left out as a validation set for model selection. The AdamW optimizer was adopted, along with the OneCycle scheduler, to optimize the model weights. Results: Our murmur detection classifier received a weighted accuracy score of 0.736 (ranked 14th out of 40 teams) and a Challenge cost score of 12944 (ranked 19th out of 39 teams) on the hidden validation set. Conclusion: We provided a practical solution to the problems of detecting heart murmurs and providing clinical diagnosis suggestions from PCGs.
comment: 4 pages, 5 figures, Computing in Cardiology 2022, URL: https://github.com/DeepPSP/cinc2022
☆ Learning multi-scale local conditional probability models of images
Deep neural networks can learn powerful prior probability models for images, as evidenced by the high-quality generations obtained with recent score-based diffusion methods. But the means by which these networks capture complex global statistical structure, apparently without suffering from the curse of dimensionality, remain a mystery. To study this, we incorporate diffusion methods into a multi-scale decomposition, reducing dimensionality by assuming a stationary local Markov model for wavelet coefficients conditioned on coarser-scale coefficients. We instantiate this model using convolutional neural networks (CNNs) with local receptive fields, which enforce both the stationarity and Markov properties. Global structures are captured using a CNN with receptive fields covering the entire (but small) low-pass image. We test this model on a dataset of face images, which are highly non-stationary and contain large-scale geometric structures. Remarkably, denoising, super-resolution, and image synthesis results all demonstrate that these structures can be captured with significantly smaller conditioning neighborhoods than required by a Markov model implemented in the pixel domain. Our results show that score estimation for large complex images can be reduced to low-dimensional Markov conditional models across scales, alleviating the curse of dimensionality.
comment: 16 pages, 8 figures
☆ KDSM: An uplift modeling framework based on knowledge distillation and sample matching
Uplift modeling aims to estimate the treatment effect on individuals, widely applied in the e-commerce platform to target persuadable customers and maximize the return of marketing activities. Among the existing uplift modeling methods, tree-based methods are adept at fitting increment and generalization, while neural-network-based models excel at predicting absolute value and precision, and these advantages have not been fully explored and combined. Also, the lack of counterfactual sample pairs is the root challenge in uplift modeling. In this paper, we proposed an uplift modeling framework based on Knowledge Distillation and Sample Matching (KDSM). The teacher model is the uplift decision tree (UpliftDT), whose structure is exploited to construct counterfactual sample pairs, and the pairwise incremental prediction is treated as another objective for the student model. Under the idea of multitask learning, the student model can achieve better performance on generalization and even surpass the teacher. Extensive offline experiments validate the universality of different combinations of teachers and student models and the superiority of KDSM measured against the baselines. In online A/B testing, the cost of each incremental room night is reduced by 6.5\%.
☆ Rethinking Confidence Calibration for Failure Prediction ECCV 2022
Reliable confidence estimation for the predictions is important in many safety-critical applications. However, modern deep neural networks are often overconfident for their incorrect predictions. Recently, many calibration methods have been proposed to alleviate the overconfidence problem. With calibrated confidence, a primary and practical purpose is to detect misclassification errors by filtering out low-confidence predictions (known as failure prediction). In this paper, we find a general, widely-existed but actually-neglected phenomenon that most confidence calibration methods are useless or harmful for failure prediction. We investigate this problem and reveal that popular confidence calibration methods often lead to worse confidence separation between correct and incorrect samples, making it more difficult to decide whether to trust a prediction or not. Finally, inspired by the natural connection between flat minima and confidence separation, we propose a simple hypothesis: flat minima is beneficial for failure prediction. We verify this hypothesis via extensive experiments and further boost the performance by combining two different flat minima techniques. Our code is available at https://github.com/Impression2805/FMFP
comment: Accepted to ECCV 2022. Code is available at https://github.com/Impression2805/FMFP
☆ Non-Parametric Outlier Synthesis ICLR 2023
Out-of-distribution (OOD) detection is indispensable for safely deploying machine learning models in the wild. One of the key challenges is that models lack supervision signals from unknown data, and as a result, can produce overconfident predictions on OOD data. Recent work on outlier synthesis modeled the feature space as parametric Gaussian distribution, a strong and restrictive assumption that might not hold in reality. In this paper, we propose a novel framework, Non-Parametric Outlier Synthesis (NPOS), which generates artificial OOD training data and facilitates learning a reliable decision boundary between ID and OOD data. Importantly, our proposed synthesis approach does not make any distributional assumption on the ID embeddings, thereby offering strong flexibility and generality. We show that our synthesis approach can be mathematically interpreted as a rejection sampling framework. Extensive experiments show that NPOS can achieve superior OOD detection performance, outperforming the competitive rivals by a significant margin. Code is publicly available at https://github.com/deeplearning-wisc/npos.
comment: ICLR 2023
☆ Primal and Dual Analysis of Entropic Fictitious Play for Finite-sum Problems
The entropic fictitious play (EFP) is a recently proposed algorithm that minimizes the sum of a convex functional and entropy in the space of measures -- such an objective naturally arises in the optimization of a two-layer neural network in the mean-field regime. In this work, we provide a concise primal-dual analysis of EFP in the setting where the learning problem exhibits a finite-sum structure. We establish quantitative global convergence guarantees for both the continuous-time and discrete-time dynamics based on properties of a proximal Gibbs measure introduced in Nitanda et al. (2022). Furthermore, our primal-dual framework entails a memory-efficient particle-based implementation of the EFP update, and also suggests a connection to gradient boosting methods. We illustrate the efficiency of our novel implementation in experiments including neural network optimization and image synthesis.
☆ Centroid Distance Distillation for Effective Rehearsal in Continual Learning
Rehearsal, retraining on a stored small data subset of old tasks, has been proven effective in solving catastrophic forgetting in continual learning. However, due to the sampled data may have a large bias towards the original dataset, retraining them is susceptible to driving continual domain drift of old tasks in feature space, resulting in forgetting. In this paper, we focus on tackling the continual domain drift problem with centroid distance distillation. First, we propose a centroid caching mechanism for sampling data points based on constructed centroids to reduce the sample bias in rehearsal. Then, we present a centroid distance distillation that only stores the centroid distance to reduce the continual domain drift. The experiments on four continual learning datasets show the superiority of the proposed method, and the continual domain drift can be reduced.
☆ Graph Positional Encoding via Random Feature Propagation
Two main families of node feature augmentation schemes have been explored for enhancing GNNs: random features and spectral positional encoding. Surprisingly, however, there is still no clear understanding of the relation between these two augmentation schemes. Here we propose a novel family of positional encoding schemes which draws a link between the above two approaches and improves over both. The new approach, named Random Feature Propagation (RFP), is inspired by the power iteration method and its generalizations. It concatenates several intermediate steps of an iterative algorithm for computing the dominant eigenvectors of a propagation matrix, starting from random node features. Notably, these propagation steps are based on graph-dependent propagation operators that can be either predefined or learned. We explore the theoretical and empirical benefits of RFP. First, we provide theoretical justifications for using random features, for incorporating early propagation steps, and for using multiple random initializations. Then, we empirically demonstrate that RFP significantly outperforms both spectral PE and random features in multiple node classification and graph classification benchmarks.
☆ A Topological Distance Measure between Multi-Fields for Classification and Analysis of Shapes and Data
Distance measures play an important role in shape classification and data analysis problems. Topological distances based on Reeb graphs and persistence diagrams have been employed to obtain effective algorithms in shape matching and scalar data analysis. In the current paper, we propose an improved distance measure between two multi-fields by computing a multi-dimensional Reeb graph (MDRG) each of which captures the topology of a multi-field through a hierarchy of Reeb graphs in different dimensions. A hierarchy of persistence diagrams is then constructed by computing a persistence diagram corresponding to each Reeb graph of the MDRG. Based on this representation, we propose a novel distance measure between two MDRGs by extending the bottleneck distance between two Reeb graphs. We show that the proposed measure satisfies the pseudo-metric and stability properties. We examine the effectiveness of the proposed multi-field topology-based measure on two different applications: (1) shape classification and (2) detection of topological features in a time-varying multi-field data. In the shape classification problem, the performance of the proposed measure is compared with the well-known topology-based measures in shape matching. In the second application, we consider a time-varying volumetric multi-field data from the field of computational chemistry where the goal is to detect the site of stable bond formation between Pt and CO molecules. We demonstrate the ability of the proposed distance in classifying each of the sites as occurring before and after the bond stabilization.
☆ The $α$-divergence Improves the Entropy Production Estimation via Machine Learning
Recent years have seen a surge of interest in the algorithmic estimation of stochastic entropy production (EP) from the trajectory data via machine learning. A crucial element of such algorithms is the identification of a loss function whose minimization guarantees the accurate EP estimation. In this study, we show that there exists a host of loss functions, namely those implementing a variational representation of the $\alpha$-divergence, which can be used for the EP estimation. Among these loss functions, the one corresponding to $\alpha = -0.5$ exhibits the most robust performance against strong nonequilibrium driving or slow dynamics, which adversely affects the existing method based on the Kullback-Leibler divergence ($\alpha = 0$). To corroborate our findings, we present an exactly solvable simplification of the EP estimation problem, whose loss function landscape and stochastic properties demonstrate the optimality of $\alpha = -0.5$.
comment: 9 pages, 7 figures
☆ Stabilizing the Maximal Entropy Moment Method for Rarefied Gas Dynamics at Single-Precision
Developing extended hydrodynamics equations valid for both dense and rarefied gases remains a great challenge. A systematical solution for this challenge is the moment method describing both dense and rarefied gas behaviors with moments of gas molecule velocity distributions. Among moment methods, the maximal entropy moment method (MEM) stands out for its well-posedness and stability, which utilizes velocity distributions with maximized entropy. However, finding such distributions requires solving an ill-conditioned and computation-demanding optimization problem. This problem causes numerical overflow and breakdown when the numerical precision is insufficient, especially for flows like high-speed shock waves. It also prevents modern GPUs from accelerating optimization with their enormous single floating-point precision computation power. This paper aims to stabilize MEM, making it practical for simulating very strong normal shock waves on modern GPUs at single precision. We propose the gauge transformations for MEM, making the optimization less ill-conditioned. We also tackle numerical overflow and breakdown by adopting the canonical form of distribution and Newton's modified optimization method. With these techniques, we achieved a single-precision GPU simulation of a Mach 10 shock wave with 35 moments MEM, surpassing the previous double-precision results of Mach 4. Moreover, we argued that over-refined spatial mesh degrades both the accuracy and stability of MEM. Overall, this paper makes the maximal entropy moment method practical for simulating very strong normal shock waves on modern GPUs at single-precision, with significant stability improvement compared to previous methods.
comment: 25 pages, 5 figures
☆ Perspectives on the Social Impacts of Reinforcement Learning with Human Feedback
Is it possible for machines to think like humans? And if it is, how should we go about teaching them to do so? As early as 1950, Alan Turing stated that we ought to teach machines in the way of teaching a child. Reinforcement learning with human feedback (RLHF) has emerged as a strong candidate toward allowing agents to learn from human feedback in a naturalistic manner. RLHF is distinct from traditional reinforcement learning as it provides feedback from a human teacher in addition to a reward signal. It has been catapulted into public view by multiple high-profile AI applications, including OpenAI's ChatGPT, DeepMind's Sparrow, and Anthropic's Claude. These highly capable chatbots are already overturning our understanding of how AI interacts with humanity. The wide applicability and burgeoning success of RLHF strongly motivate the need to evaluate its social impacts. In light of recent developments, this paper considers an important question: can RLHF be developed and used without negatively affecting human societies? Our objectives are threefold: to provide a systematic study of the social effects of RLHF; to identify key social and ethical issues of RLHF; and to discuss social impacts for stakeholders. Although text-based applications of RLHF have received much attention, it is crucial to consider when evaluating its social implications the diverse range of areas to which it may be deployed. We describe seven primary ways in which RLHF-based technologies will affect society by positively transforming human experiences with AI. This paper ultimately proposes that RLHF has potential to net positively impact areas of misinformation, AI value-alignment, bias, AI access, cross-cultural dialogue, industry, and workforce. As RLHF raises concerns that echo those of existing AI technologies, it will be important for all to be aware and intentional in the adoption of RLHF.
☆ An Analysis of Physics-Informed Neural Networks
Whilst the partial differential equations that govern the dynamics of our world have been studied in great depth for centuries, solving them for complex, high-dimensional conditions and domains still presents an incredibly large mathematical and computational challenge. Analytical methods can be cumbersome to utilise, and numerical methods can lead to errors and inaccuracies. On top of this, sometimes we lack the information or knowledge to pose the problem well enough to apply these kinds of methods. Here, we present a new approach to approximating the solution to physical systems - physics-informed neural networks. The concept of artificial neural networks is introduced, the objective function is defined, and optimisation strategies are discussed. The partial differential equation is then included as a constraint in the loss function for the optimisation problem, giving the network access to knowledge of the dynamics of the physical system it is modelling. Some intuitive examples are displayed, and more complex applications are considered to showcase the power of physics informed neural networks, such as in seismic imaging. Solution error is analysed, and suggestions are made to improve convergence and/or solution precision. Problems and limitations are also touched upon in the conclusions, as well as some thoughts as to where physics informed neural networks are most useful, and where they could go next.
comment: 56 pages
☆ Model Sketching: Centering Concepts in Early-Stage Machine Learning Model Design
Machine learning practitioners often end up tunneling on low-level technical details like model architectures and performance metrics. Could early model development instead focus on high-level questions of which factors a model ought to pay attention to? Inspired by the practice of sketching in design, which distills ideas to their minimal representation, we introduce model sketching: a technical framework for iteratively and rapidly authoring functional approximations of a machine learning model's decision-making logic. Model sketching refocuses practitioner attention on composing high-level, human-understandable concepts that the model is expected to reason over (e.g., profanity, racism, or sarcasm in a content moderation task) using zero-shot concept instantiation. In an evaluation with 17 ML practitioners, model sketching reframed thinking from implementation to higher-level exploration, prompted iteration on a broader range of model designs, and helped identify gaps in the problem formulation$\unicode{x2014}$all in a fraction of the time ordinarily required to build a model.
comment: To appear at CHI 2023
☆ Very fast, approximate counterfactual explanations for decision forests AAAI 2023
We consider finding a counterfactual explanation for a classification or regression forest, such as a random forest. This requires solving an optimization problem to find the closest input instance to a given instance for which the forest outputs a desired value. Finding an exact solution has a cost that is exponential on the number of leaves in the forest. We propose a simple but very effective approach: we constrain the optimization to only those input space regions defined by the forest that are populated by actual data points. The problem reduces to a form of nearest-neighbor search using a certain distance on a certain dataset. This has two advantages: first, the solution can be found very quickly, scaling to large forests and high-dimensional data, and enabling interactive use. Second, the solution found is more likely to be realistic in that it is guided towards high-density areas of input space.
comment: A shorter version of this paper appears in AAAI 2023
☆ Spatiotemporal Capsule Neural Network for Vehicle Trajectory Prediction
Through advancement of the Vehicle-to-Everything (V2X) network, road safety, energy consumption, and traffic efficiency can be significantly improved. An accurate vehicle trajectory prediction benefits communication traffic management and network resource allocation for the real-time application of the V2X network. Recurrent neural networks and their variants have been reported in recent research to predict vehicle mobility. However, the spatial attribute of vehicle movement behavior has been overlooked, resulting in incomplete information utilization. To bridge this gap, we put forward for the first time a hierarchical trajectory prediction structure using the capsule neural network (CapsNet) with three sequential components. First, the geographic information is transformed into a grid map presentation, describing vehicle mobility distribution spatially and temporally. Second, CapsNet serves as the core model to embed local temporal and global spatial correlation through hierarchical capsules. Finally, extensive experiments conducted on actual taxi mobility data collected in Porto city (Portugal) and Singapore show that the proposed method outperforms the state-of-the-art methods.
comment: IEEE TVT has accepted this paper
☆ A Review of Deep Learning-Powered Mesh Reconstruction Methods
With the recent advances in hardware and rendering techniques, 3D models have emerged everywhere in our life. Yet creating 3D shapes is arduous and requires significant professional knowledge. Meanwhile, Deep learning has enabled high-quality 3D shape reconstruction from various sources, making it a viable approach to acquiring 3D shapes with minimal effort. Importantly, to be used in common 3D applications, the reconstructed shapes need to be represented as polygonal meshes, which is a challenge for neural networks due to the irregularity of mesh tessellations. In this survey, we provide a comprehensive review of mesh reconstruction methods that are powered by machine learning. We first describe various representations for 3D shapes in the deep learning context. Then we review the development of 3D mesh reconstruction methods from voxels, point clouds, single images, and multi-view images. Finally, we identify several challenges in this field and propose potential future directions.
☆ Finding metastable skyrmionic structures via a metaheuristic perturbation-driven neural network
Topological magnetic textures observed in experiments can, in principle, be predicted by theoretical calculations and numerical simulations. However, such calculations are, in general, hampered by difficulties in distinguishing between local and global energy minima. This becomes particularly problematic for magnetic materials that allow for a multitude of topological charges. Finding solutions to such problems by means of classical numerical methods can be challenging because either a good initial guess or a gigantic amount of random sampling is required. In this study, we demonstrate an efficient way to identify those metastable configurations by leveraging the power of gradient descent-based optimization within the framework of a feedforward neural network combined with a heuristic meta-search, which is driven by a random perturbation of the neural network's input. We exemplify the power of the method by an analysis of the Pd/Fe/Ir(111) system, an experimentally well characterized system.
☆ DR-Label: Improving GNN Models for Catalysis Systems by Label Deconstruction and Reconstruction
Attaining the equilibrium state of a catalyst-adsorbate system is key to fundamentally assessing its effective properties, such as adsorption energy. Machine learning methods with finer supervision strategies have been applied to boost and guide the relaxation process of an atomic system and better predict its properties at the equilibrium state. In this paper, we present a novel graph neural network (GNN) supervision and prediction strategy DR-Label. The method enhances the supervision signal, reduces the multiplicity of solutions in edge representation, and encourages the model to provide node predictions that are graph structural variation robust. DR-Label first Deconstructs finer-grained equilibrium state information to the model by projecting the node-level supervision signal to each edge. Reversely, the model Reconstructs a more robust equilibrium state prediction by transforming edge-level predictions to node-level with a sphere-fitting algorithm. The DR-Label strategy was applied to three radically distinct models, each of which displayed consistent performance enhancements. Based on the DR-Label strategy, we further proposed DRFormer, which achieved a new state-of-the-art performance on the Open Catalyst 2020 (OC20) dataset and the Cu-based single-atom-alloyed CO adsorption (SAA) dataset. We expect that our work will highlight crucial steps for the development of a more accurate model in equilibrium state property prediction of a catalysis system.
comment: 11 pages, 3 figures
☆ Adversarial Sampling for Fairness Testing in Deep Neural Network
In this research, we focus on the usage of adversarial sampling to test for the fairness in the prediction of deep neural network model across different classes of image in a given dataset. While several framework had been proposed to ensure robustness of machine learning model against adversarial attack, some of which includes adversarial training algorithm. There is still the pitfall that adversarial training algorithm tends to cause disparity in accuracy and robustness among different group. Our research is aimed at using adversarial sampling to test for fairness in the prediction of deep neural network model across different classes or categories of image in a given dataset. We successfully demonstrated a new method of ensuring fairness across various group of input in deep neural network classifier. We trained our neural network model on the original image, and without training our model on the perturbed or attacked image. When we feed the adversarial samplings to our model, it was able to predict the original category/ class of the image the adversarial sample belongs to. We also introduced and used the separation of concern concept from software engineering whereby there is an additional standalone filter layer that filters perturbed image by heavily removing the noise or attack before automatically passing it to the network for classification, we were able to have accuracy of 93.3%. Cifar-10 dataset have ten categories of dataset, and so, in order to account for fairness, we applied our hypothesis across each categories of dataset and were able to get a consistent result and accuracy.
comment: 7 pages, 5 figures, International Journal of Advanced Computer Science and Application
☆ Angel-PTM: A Scalable and Economical Large-scale Pre-training System in Tencent
Recent years have witnessed the unprecedented achievements of large-scale pre-trained models, especially the Transformer models. Many products and services in Tencent Inc., such as WeChat, QQ, and Tencent Advertisement, have been opted in to gain the power of pre-trained models. In this work, we present Angel-PTM, a productive deep learning system designed for pre-training and fine-tuning Transformer models. Angel-PTM can train extremely large-scale models with hierarchical memory efficiently. The key designs of Angel-PTM are the fine-grained memory management via the Page abstraction and a unified scheduling method that coordinate the computations, data movements, and communications. Furthermore, Angel-PTM supports extreme model scaling with SSD storage and implements the lock-free updating mechanism to address the SSD I/O bandwidth bottlenecks. Experimental results demonstrate that Angel-PTM outperforms existing systems by up to 114.8% in terms of maximum model scale as well as up to 88.9% in terms of training throughput. Additionally, experiments on GPT3-175B and T5-MoE-1.2T models utilizing hundreds of GPUs verify the strong scalability of Angel-PTM.
☆ A Multi-Grained Self-Interpretable Symbolic-Neural Model For Single/Multi-Labeled Text Classification ICLR 2023
Deep neural networks based on layer-stacking architectures have historically suffered from poor inherent interpretability. Meanwhile, symbolic probabilistic models function with clear interpretability, but how to combine them with neural networks to enhance their performance remains to be explored. In this paper, we try to marry these two systems for text classification via a structured language model. We propose a Symbolic-Neural model that can learn to explicitly predict class labels of text spans from a constituency tree without requiring any access to span-level gold labels. As the structured language model learns to predict constituency trees in a self-supervised manner, only raw texts and sentence-level labels are required as training data, which makes it essentially a general constituent-level self-interpretable classification model. Our experiments demonstrate that our approach could achieve good prediction accuracy in downstream tasks. Meanwhile, the predicted span labels are consistent with human rationales to a certain degree.
comment: ICLR 2023
☆ Bayesian inference with finitely wide neural networks
The analytic inference, e.g. predictive distribution being in closed form, may be an appealing benefit for machine learning practitioners when they treat wide neural networks as Gaussian process in Bayesian setting. The realistic widths, however, are finite and cause weak deviation from the Gaussianity under which partial marginalization of random variables in a model is straightforward. On the basis of multivariate Edgeworth expansion, we propose a non-Gaussian distribution in differential form to model a finite set of outputs from a random neural network, and derive the corresponding marginal and conditional properties. Thus, we are able to derive the non-Gaussian posterior distribution in Bayesian regression task. In addition, in the bottlenecked deep neural networks, a weight space representation of deep Gaussian process, the non-Gaussianity is investigated through the marginal kernel.
comment: Preprint
☆ Knowledge-embedded meta-learning model for lift coefficient prediction of airfoils
Aerodynamic performance evaluation is an important part of the aircraft aerodynamic design optimization process; however, traditional methods are costly and time-consuming. Despite the fact that various machine learning methods can achieve high accuracy, their application in engineering is still difficult due to their poor generalization performance and "black box" nature. In this paper, a knowledge-embedded meta learning model, which fully integrates data with the theoretical knowledge of the lift curve, is developed to obtain the lift coefficients of an arbitrary supercritical airfoil under various angle of attacks. In the proposed model, a primary network is responsible for representing the relationship between the lift and angle of attack, while the geometry information is encoded into a hyper network to predict the unknown parameters involved in the primary network. Specifically, three models with different architectures are trained to provide various interpretations. Compared to the ordinary neural network, our proposed model can exhibit better generalization capability with competitive prediction accuracy. Afterward, interpretable analysis is performed based on the Integrated Gradients and Saliency methods. Results show that the proposed model can tend to assess the influence of airfoil geometry to the physical characteristics. Furthermore, the exceptions and shortcomings caused by the proposed model are analysed and discussed in detail.
☆ Model-Agnostic Meta-Learning for Natural Language Understanding Tasks in Finance
Natural language understanding(NLU) is challenging for finance due to the lack of annotated data and the specialized language in that domain. As a result, researchers have proposed to use pre-trained language model and multi-task learning to learn robust representations. However, aggressive fine-tuning often causes over-fitting and multi-task learning may favor tasks with significantly larger amounts data, etc. To address these problems, in this paper, we investigate model-agnostic meta-learning algorithm(MAML) in low-resource financial NLU tasks. Our contribution includes: 1. we explore the performance of MAML method with multiple types of tasks: GLUE datasets, SNLI, Sci-Tail and Financial PhraseBank; 2. we study the performance of MAML method with multiple single-type tasks: a real scenario stock price prediction problem with twitter text data. Our models achieve the state-of-the-art performance according to the experimental results, which demonstrate that our method can adapt fast and well to low-resource situations.
comment: 13 pages, 6 figures, 8 tables
☆ eCDANs: Efficient Temporal Causal Discovery from Autocorrelated and Non-stationary Data (Student Abstract)
Conventional temporal causal discovery (CD) methods suffer from high dimensionality, fail to identify lagged causal relationships, and often ignore dynamics in relations. In this study, we present a novel constraint-based CD approach for autocorrelated and non-stationary time series data (eCDANs) capable of detecting lagged and contemporaneous causal relationships along with temporal changes. eCDANs addresses high dimensionality by optimizing the conditioning sets while conducting conditional independence (CI) tests and identifies the changes in causal relations by introducing a surrogate variable to represent time dependency. Experiments on synthetic and real-world data show that eCDANs can identify time influence and outperform the baselines.
☆ Attribution-Scores and Causal Counterfactuals as Explanations in Artificial Intelligence
In this expository article we highlight the relevance of explanations for artificial intelligence, in general, and for the newer developments in {\em explainable AI}, referring to origins and connections of and among different approaches. We describe in simple terms, explanations in data management and machine learning that are based on attribution-scores, and counterfactuals as found in the area of causality. We elaborate on the importance of logical reasoning when dealing with counterfactuals, and their use for score computation.
comment: Submitted as chapter contribution
☆ Robust Autoencoders for Collective Corruption Removal ICASSP2023
Robust PCA is a standard tool for learning a linear subspace in the presence of sparse corruption or rare outliers. What about robustly learning manifolds that are more realistic models for natural data, such as images? There have been several recent attempts to generalize robust PCA to manifold settings. In this paper, we propose $\ell_1$- and scaling-invariant $\ell_1/\ell_2$-robust autoencoders based on a surprisingly compact formulation built on the intuition that deep autoencoders perform manifold learning. We demonstrate on several standard image datasets that the proposed formulation significantly outperforms all previous methods in collectively removing sparse corruption, without clean images for training. Moreover, we also show that the learned manifold structures can be generalized to unseen data samples effectively.
comment: This paper has been accepted to ICASSP2023
☆ Visual Analytics of Neuron Vulnerability to Adversarial Attacks on Convolutional Neural Networks
Adversarial attacks on a convolutional neural network (CNN) -- injecting human-imperceptible perturbations into an input image -- could fool a high-performance CNN into making incorrect predictions. The success of adversarial attacks raises serious concerns about the robustness of CNNs, and prevents them from being used in safety-critical applications, such as medical diagnosis and autonomous driving. Our work introduces a visual analytics approach to understanding adversarial attacks by answering two questions: (1) which neurons are more vulnerable to attacks and (2) which image features do these vulnerable neurons capture during the prediction? For the first question, we introduce multiple perturbation-based measures to break down the attacking magnitude into individual CNN neurons and rank the neurons by their vulnerability levels. For the second, we identify image features (e.g., cat ears) that highly stimulate a user-selected neuron to augment and validate the neuron's responsibility. Furthermore, we support an interactive exploration of a large number of neurons by aiding with hierarchical clustering based on the neurons' roles in the prediction. To this end, a visual analytics system is designed to incorporate visual reasoning for interpreting adversarial attacks. We validate the effectiveness of our system through multiple case studies as well as feedback from domain experts.
comment: Accepted by the Special Issue on Human-Centered Explainable AI, ACM Transactions on Interactive Intelligent Systems
☆ Robust Dominant Periodicity Detection for Time Series with Missing Data ICASSP 2023
Periodicity detection is an important task in time series analysis, but still a challenging problem due to the diverse characteristics of time series data like abrupt trend change, outlier, noise, and especially block missing data. In this paper, we propose a robust and effective periodicity detection algorithm for time series with block missing data. We first design a robust trend filter to remove the interference of complicated trend patterns under missing data. Then, we propose a robust autocorrelation function (ACF) that can handle missing values and outliers effectively. We rigorously prove that the proposed robust ACF can still work well when the length of the missing block is less than $1/3$ of the period length. Last, by combining the time-frequency information, our algorithm can generate the period length accurately. The experimental results demonstrate that our algorithm outperforms existing periodicity detection algorithms on real-world time series datasets.
comment: Accepted by 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP 2023)
☆ Large Language Models as Zero-Shot Human Models for Human-Robot Interaction
Human models play a crucial role in human-robot interaction (HRI), enabling robots to consider the impact of their actions on people and plan their behavior accordingly. However, crafting good human models is challenging; capturing context-dependent human behavior requires significant prior knowledge and/or large amounts of interaction data, both of which are difficult to obtain. In this work, we explore the potential of large-language models (LLMs) -- which have consumed vast amounts of human-generated text data -- to act as zero-shot human models for HRI. Our experiments on three social datasets yield promising results; the LLMs are able to achieve performance comparable to purpose-built models. That said, we also discuss current limitations, such as sensitivity to prompts and spatial/numerical reasoning mishaps. Based on our findings, we demonstrate how LLM-based human models can be integrated into a social robot's planning process and applied in HRI scenarios. Specifically, we present one case study on a simulated trust-based table-clearing task and replicate past results that relied on custom models. Next, we conduct a new robot utensil-passing experiment (n = 65) where preliminary results show that planning with a LLM-based human model can achieve gains over a basic myopic plan. In summary, our results show that LLMs offer a promising (but incomplete) approach to human modeling for HRI.
comment: 8 pages
☆ Expressivity of Shallow and Deep Neural Networks for Polynomial Approximation
We analyze the number of neurons that a ReLU neural network needs to approximate multivariate monomials. We establish an exponential lower bound for the complexity of any shallow network that approximates the product function $\vec{x} \to \prod_{i=1}^d x_i$ on a general compact domain. Furthermore, we prove that this lower bound does not hold for normalized O(1)-Lipschitz monomials (or equivalently, by restricting to the unit cube). These results suggest shallow ReLU networks suffer from the curse of dimensionality when expressing functions with a Lipschitz parameter scaling with the dimension of the input, and that the expressive power of neural networks lies in their depth rather than the overall complexity.
comment: 14 pages
☆ 3D Equivariant Diffusion for Target-Aware Molecule Generation and Affinity Prediction ICLR 2023
Rich data and powerful machine learning models allow us to design drugs for a specific protein target \textit{in silico}. Recently, the inclusion of 3D structures during targeted drug design shows superior performance to other target-free models as the atomic interaction in the 3D space is explicitly modeled. However, current 3D target-aware models either rely on the voxelized atom densities or the autoregressive sampling process, which are not equivariant to rotation or easily violate geometric constraints resulting in unrealistic structures. In this work, we develop a 3D equivariant diffusion model to solve the above challenges. To achieve target-aware molecule design, our method learns a joint generative process of both continuous atom coordinates and categorical atom types with a SE(3)-equivariant network. Moreover, we show that our model can serve as an unsupervised feature extractor to estimate the binding affinity under proper parameterization, which provides an effective way for drug screening. To evaluate our model, we propose a comprehensive framework to evaluate the quality of sampled molecules from different dimensions. Empirical studies show our model could generate molecules with more realistic 3D structures and better affinities towards the protein targets, and improve binding affinity ranking and prediction without retraining.
comment: Accepted to ICLR 2023
♻ ☆ Score-based Continuous-time Discrete Diffusion Models
Score-based modeling through stochastic differential equations (SDEs) has provided a new perspective on diffusion models, and demonstrated superior performance on continuous data. However, the gradient of the log-likelihood function, i.e., the score function, is not properly defined for discrete spaces. This makes it non-trivial to adapt \textcolor{\cdiff}{the score-based modeling} to categorical data. In this paper, we extend diffusion models to discrete variables by introducing a stochastic jump process where the reverse process denoises via a continuous-time Markov chain. This formulation admits an analytical simulation during backward sampling. To learn the reverse process, we extend score matching to general categorical data and show that an unbiased estimator can be obtained via simple matching of the conditional marginal distributions. We demonstrate the effectiveness of the proposed method on a set of synthetic and real-world music and image benchmarks.
♻ ☆ Efficient Domain Coverage for Vehicles with Second-Order Dynamics via Multi-Agent Reinforcement Learning
Collaborative autonomous multi-agent systems covering a specified area have many potential applications, such as UAV search and rescue, forest fire fighting, and real-time high-resolution monitoring. Traditional approaches for such coverage problems involve designing a model-based control policy based on sensor data. However, designing model-based controllers is challenging, and the state-of-the-art classical control policy still exhibits a large degree of sub-optimality. In this paper, we present a reinforcement learning (RL) approach for the multi-agent efficient domain coverage problem involving agents with second-order dynamics. Our approach is based on the Multi-Agent Proximal Policy Optimization Algorithm (MAPPO). Our proposed network architecture includes the incorporation of LSTM and self-attention, which allows the trained policy to adapt to a variable number of agents. Our trained policy significantly outperforms the state-of-the-art classical control policy. We demonstrate our proposed method in a variety of simulated experiments.
comment: This version is submitted to IEEE / RSJ International Conference on Intelligent Robots and Systems, 2023
♻ ☆ Transformer Meets Boundary Value Inverse Problems ICLR 2023
A Transformer-based deep direct sampling method is proposed for electrical impedance tomography, a well-known severely ill-posed nonlinear boundary value inverse problem. A real-time reconstruction is achieved by evaluating the learned inverse operator between carefully designed data and the reconstructed images. An effort is made to give a specific example to a fundamental question: whether and how one can benefit from the theoretical structure of a mathematical problem to develop task-oriented and structure-conforming deep neural networks? Specifically, inspired by direct sampling methods for inverse problems, the 1D boundary data in different frequencies are preprocessed by a partial differential equation-based feature map to yield 2D harmonic extensions as different input channels. Then, by introducing learnable non-local kernels, the direct sampling is recast to a modified attention mechanism. The new method achieves superior accuracy over its predecessors and contemporary operator learners and shows robustness to noises in benchmarks. This research shall strengthen the insights that, despite being invented for natural language processing tasks, the attention mechanism offers great flexibility to be modified in conformity with the a priori mathematical knowledge, which ultimately leads to the design of more physics-compatible neural architectures.
comment: 30 pages, 10 figures. Published as a conference paper at ICLR 2023
♻ ☆ All are Worth Words: A ViT Backbone for Diffusion Models
Vision transformers (ViT) have shown promise in various vision tasks while the U-Net based on a convolutional neural network (CNN) remains dominant in diffusion models. We design a simple and general ViT-based architecture (named U-ViT) for image generation with diffusion models. U-ViT is characterized by treating all inputs including the time, condition and noisy image patches as tokens and employing long skip connections between shallow and deep layers. We evaluate U-ViT in unconditional and class-conditional image generation, as well as text-to-image generation tasks, where U-ViT is comparable if not superior to a CNN-based U-Net of a similar size. In particular, latent diffusion models with U-ViT achieve record-breaking FID scores of 2.29 in class-conditional image generation on ImageNet 256x256, and 5.48 in text-to-image generation on MS-COCO, among methods without accessing large external datasets during the training of generative models. Our results suggest that, for diffusion-based image modeling, the long skip connection is crucial while the down-sampling and up-sampling operators in CNN-based U-Net are not always necessary. We believe that U-ViT can provide insights for future research on backbones in diffusion models and benefit generative modeling on large scale cross-modality datasets.
♻ ☆ Generating a Terrain-Robustness Benchmark for Legged Locomotion: A Prototype via Terrain Authoring and Active Learning ICRA 2023
Terrain-aware locomotion has become an emerging topic in legged robotics. However, it is hard to generate diverse, challenging, and realistic unstructured terrains in simulation, which limits the way researchers evaluate their locomotion policies. In this paper, we prototype the generation of a terrain dataset via terrain authoring and active learning, and the learned samplers can stably generate diverse high-quality terrains. We expect the generated dataset to make a terrain-robustness benchmark for legged locomotion. The dataset, the code implementation, and some policy evaluations are released at https://bit.ly/3bn4j7f.
comment: 7 pages, 7 figures. IEEE ICRA 2023
♻ ☆ Robust Multivariate Time-Series Forecasting: Adversarial Attacks and Defense Mechanisms
This work studies the threats of adversarial attack on multivariate probabilistic forecasting models and viable defense mechanisms. Our studies discover a new attack pattern that negatively impact the forecasting of a target time series via making strategic, sparse (imperceptible) modifications to the past observations of a small number of other time series. To mitigate the impact of such attack, we have developed two defense strategies. First, we extend a previously developed randomized smoothing technique in classification to multivariate forecasting scenarios. Second, we develop an adversarial training algorithm that learns to create adversarial examples and at the same time optimizes the forecasting model to improve its robustness against such adversarial simulation. Extensive experiments on real-world datasets confirm that our attack schemes are powerful and our defense algorithms are more effective compared with baseline defense mechanisms.
♻ ☆ Noise2Music: Text-conditioned Music Generation with Diffusion Models
We introduce Noise2Music, where a series of diffusion models is trained to generate high-quality 30-second music clips from text prompts. Two types of diffusion models, a generator model, which generates an intermediate representation conditioned on text, and a cascader model, which generates high-fidelity audio conditioned on the intermediate representation and possibly the text, are trained and utilized in succession to generate high-fidelity music. We explore two options for the intermediate representation, one using a spectrogram and the other using audio with lower fidelity. We find that the generated audio is not only able to faithfully reflect key elements of the text prompt such as genre, tempo, instruments, mood, and era, but goes beyond to ground fine-grained semantics of the prompt. Pretrained large language models play a key role in this story -- they are used to generate paired text for the audio of the training set and to extract embeddings of the text prompts ingested by the diffusion models. Generated examples: https://google-research.github.io/noise2music
comment: 15 pages
♻ ☆ Falsification before Extrapolation in Causal Effect Estimation
Randomized Controlled Trials (RCTs) represent a gold standard when developing policy guidelines. However, RCTs are often narrow, and lack data on broader populations of interest. Causal effects in these populations are often estimated using observational datasets, which may suffer from unobserved confounding and selection bias. Given a set of observational estimates (e.g. from multiple studies), we propose a meta-algorithm that attempts to reject observational estimates that are biased. We do so using validation effects, causal effects that can be inferred from both RCT and observational data. After rejecting estimators that do not pass this test, we generate conservative confidence intervals on the extrapolated causal effects for subgroups not observed in the RCT. Under the assumption that at least one observational estimator is asymptotically normal and consistent for both the validation and extrapolated effects, we provide guarantees on the coverage probability of the intervals output by our algorithm. To facilitate hypothesis testing in settings where causal effect transportation across datasets is necessary, we give conditions under which a doubly-robust estimator of group average treatment effects is asymptotically normal, even when flexible machine learning methods are used for estimation of nuisance parameters. We illustrate the properties of our approach on semi-synthetic and real world datasets, and show that it compares favorably to standard meta-analysis techniques.
comment: Conference on Neural Information Processing Systems, 2022
♻ ☆ Toward Certified Robustness Against Real-World Distribution Shifts
We consider the problem of certifying the robustness of deep neural networks against real-world distribution shifts. To do so, we bridge the gap between hand-crafted specifications and realistic deployment settings by proposing a novel neural-symbolic verification framework, in which we train a generative model to learn perturbations from data and define specifications with respect to the output of the learned model. A unique challenge arising from this setting is that existing verifiers cannot tightly approximate sigmoid activations, which are fundamental to many state-of-the-art generative models. To address this challenge, we propose a general meta-algorithm for handling sigmoid activations which leverages classical notions of counter-example-guided abstraction refinement. The key idea is to "lazily" refine the abstraction of sigmoid functions to exclude spurious counter-examples found in the previous abstraction, thus guaranteeing progress in the verification process while keeping the state-space small. Experiments on the MNIST and CIFAR-10 datasets show that our framework significantly outperforms existing methods on a range of challenging distribution shifts.
comment: SatML'23. Keywords: certified robustness, distribution shift, generative models, S-shaped activations, CEGAR
♻ ☆ Cross-domain Transfer Learning and State Inference for Soft Robots via a Semi-supervised Sequential Variational Bayes Framework ICRA
Recently, data-driven models such as deep neural networks have shown to be promising tools for modelling and state inference in soft robots. However, voluminous amounts of data are necessary for deep models to perform effectively, which requires exhaustive and quality data collection, particularly of state labels. Consequently, obtaining labelled state data for soft robotic systems is challenged for various reasons, including difficulty in the sensorization of soft robots and the inconvenience of collecting data in unstructured environments. To address this challenge, in this paper, we propose a semi-supervised sequential variational Bayes (DSVB) framework for transfer learning and state inference in soft robots with missing state labels on certain robot configurations. Considering that soft robots may exhibit distinct dynamics under different robot configurations, a feature space transfer strategy is also incorporated to promote the adaptation of latent features across multiple configurations. Unlike existing transfer learning approaches, our proposed DSVB employs a recurrent neural network to model the nonlinear dynamics and temporal coherence in soft robot data. The proposed framework is validated on multiple setup configurations of a pneumatic-based soft robot finger. Experimental results on four transfer scenarios demonstrate that DSVB performs effective transfer learning and accurate state inference amidst missing state labels.
comment: Accepted at the International Conference on Robotics and Automation (ICRA) 2023
♻ ☆ RRWaveNet: A Compact End-to-End Multi-Scale Residual CNN for Robust PPG Respiratory Rate Estimation
Respiratory rate (RR) is an important biomarker as RR changes can reflect severe medical events such as heart disease, lung disease, and sleep disorders. Unfortunately, standard manual RR counting is prone to human error and cannot be performed continuously. This study proposes a method for continuously estimating RR, RRWaveNet. The method is a compact end-to-end deep learning model which does not require feature engineering and can use low-cost raw photoplethysmography (PPG) as input signal. RRWaveNet was tested subject-independently and compared to baseline in four datasets (BIDMC, CapnoBase, WESAD, and SensAI) and using three window sizes (16, 32, and 64 seconds). RRWaveNet outperformed current state-of-the-art methods with mean absolute errors at optimal window size of 1.66 \pm 1.01, 1.59 \pm 1.08, 1.92 \pm 0.96 and 1.23 \pm 0.61 breaths per minute for each dataset. In remote monitoring settings, such as in the WESAD and SensAI datasets, we apply transfer learning to improve the performance using two other ICU datasets as pretraining datasets, reducing the MAE by up to 21$\%$. This shows that this model allows accurate and practical estimation of RR on affordable and wearable devices. Our study also shows feasibility of remote RR monitoring in the context of telemedicine and at home.
comment: 11 pages, 8 figures
♻ ☆ Falsification of Internal and External Validity in Observational Studies via Conditional Moment Restrictions
Randomized Controlled Trials (RCT)s are relied upon to assess new treatments, but suffer from limited power to guide personalized treatment decisions. On the other hand, observational (i.e., non-experimental) studies have large and diverse populations, but are prone to various biases (e.g. residual confounding). To safely leverage the strengths of observational studies, we focus on the problem of falsification, whereby RCTs are used to validate causal effect estimates learned from observational data. In particular, we show that, given data from both an RCT and an observational study, assumptions on internal and external validity have an observable, testable implication in the form of a set of Conditional Moment Restrictions (CMRs). Further, we show that expressing these CMRs with respect to the causal effect, or "causal contrast", as opposed to individual counterfactual means, provides a more reliable falsification test. In addition to giving guarantees on the asymptotic properties of our test, we demonstrate superior power and type I error of our approach on semi-synthetic and real world datasets. Our approach is interpretable, allowing a practitioner to visualize which subgroups in the population lead to falsification of an observational study.
comment: Artificial Intelligence and Statistics 2023
♻ ☆ Model Explanation Disparities as a Fairness Diagnostic
In recent years, there has been a flurry of research focusing on the fairness of machine learning models, and in particular on quantifying and eliminating bias against protected subgroups. One line of work generalizes the notion of protected subgroups beyond simple discrete classes by introducing the notion of a "rich subgroup", and seeks to train models that are calibrated or equalize error rates with respect to these richer subgroup classes. Largely orthogonally, local model explanation methods have been developed that given a classifier h and test point x, attribute influence for the prediction h(x) to the individual features of x. This raises a natural question: Do local model explanation methods attribute different feature importance values on average across different protected subgroups, and can we detect these disparities efficiently? If the model places high weight on a given feature in a specific protected subgroup, but not on the dataset overall (or vice versa), this could be a potential indicator of bias in the predictive model or the underlying data generating process, and is at the very least a useful diagnostic that signals the need for a domain expert to delve deeper. In this paper, we formally introduce the notion of feature importance disparity (FID) in the context of rich subgroups, design oracle-efficent algorithms to identify large FID subgroups, and conduct a thorough empirical analysis that establishes auditing for FID as an important method to investigate dataset bias. Our experiments show that across 4 datasets and 4 common feature importance methods our algorithms find (feature, subgroup) pairs that simultaneously: (i) have subgroup feature importance that is often an order of magnitude different than the importance on the dataset as a whole (ii) generalize out of sample, and (iii) yield interesting discussions about potential bias inherent in these datasets.
comment: 13 pages, 6 figures. Appendix: 8 pages, 4 figures. Replacement info: minor changes to match metadata abstract to paper abstract
♻ ☆ TalkToModel: Explaining Machine Learning Models with Interactive Natural Language Conversations
Machine Learning (ML) models are increasingly used to make critical decisions in real-world applications, yet they have become more complex, making them harder to understand. To this end, researchers have proposed several techniques to explain model predictions. However, practitioners struggle to use these explainability techniques because they often do not know which one to choose and how to interpret the results of the explanations. In this work, we address these challenges by introducing TalkToModel: an interactive dialogue system for explaining machine learning models through conversations. Specifically, TalkToModel comprises of three key components: 1) a natural language interface for engaging in conversations, making ML model explainability highly accessible, 2) a dialogue engine that adapts to any tabular model and dataset, interprets natural language, maps it to appropriate explanations, and generates text responses, and 3) an execution component that constructs the explanations. We carried out extensive quantitative and human subject evaluations of TalkToModel. Overall, we found the conversational system understands user inputs on novel datasets and models with high accuracy, demonstrating the system's capacity to generalize to new situations. In real-world evaluations with humans, 73% of healthcare workers (e.g., doctors and nurses) agreed they would use TalkToModel over baseline point-and-click systems for explainability in a disease prediction task, and 85% of ML professionals agreed TalkToModel was easier to use for computing explanations. Our findings demonstrate that TalkToModel is more effective for model explainability than existing systems, introducing a new category of explainability tools for practitioners. Code & demo released here: https://github.com/dylan-slack/TalkToModel.
comment: Pre-print; comments welcome! Reach out to dslack@uci.edu v3 update title and abstract
♻ ☆ Bootstrapping Semi-supervised Medical Image Segmentation with Anatomical-aware Contrastive Distillation
Contrastive learning has shown great promise over annotation scarcity problems in the context of medical image segmentation. Existing approaches typically assume a balanced class distribution for both labeled and unlabeled medical images. However, medical image data in reality is commonly imbalanced (i.e., multi-class label imbalance), which naturally yields blurry contours and usually incorrectly labels rare objects. Moreover, it remains unclear whether all negative samples are equally negative. In this work, we present ACTION, an Anatomical-aware ConTrastive dIstillatiON framework, for semi-supervised medical image segmentation. Specifically, we first develop an iterative contrastive distillation algorithm by softly labeling the negatives rather than binary supervision between positive and negative pairs. We also capture more semantically similar features from the randomly chosen negative set compared to the positives to enforce the diversity of the sampled data. Second, we raise a more important question: Can we really handle imbalanced samples to yield better performance? Hence, the key innovation in ACTION is to learn global semantic relationship across the entire dataset and local anatomical features among the neighbouring pixels with minimal additional memory footprint. During the training, we introduce anatomical contrast by actively sampling a sparse set of hard negative pixels, which can generate smoother segmentation boundaries and more accurate predictions. Extensive experiments across two benchmark datasets and different unlabeled settings show that ACTION significantly outperforms the current state-of-the-art semi-supervised methods.
♻ ☆ Accelerating Shapley Explanation via Contributive Cooperator Selection
Even though Shapley value provides an effective explanation for a DNN model prediction, the computation relies on the enumeration of all possible input feature coalitions, which leads to the exponentially growing complexity. To address this problem, we propose a novel method SHEAR to significantly accelerate the Shapley explanation for DNN models, where only a few coalitions of input features are involved in the computation. The selection of the feature coalitions follows our proposed Shapley chain rule to minimize the absolute error from the ground-truth Shapley values, such that the computation can be both efficient and accurate. To demonstrate the effectiveness, we comprehensively evaluate SHEAR across multiple metrics including the absolute error from the ground-truth Shapley value, the faithfulness of the explanations, and running speed. The experimental results indicate SHEAR consistently outperforms state-of-the-art baseline methods across different evaluation metrics, which demonstrates its potentials in real-world applications where the computational resource is limited.
♻ ☆ Multi-Source Survival Domain Adaptation AAAI
Survival analysis is the branch of statistics that studies the relation between the characteristics of living entities and their respective survival times, taking into account the partial information held by censored cases. A good analysis can, for example, determine whether one medical treatment for a group of patients is better than another. With the rise of machine learning, survival analysis can be modeled as learning a function that maps studied patients to their survival times. To succeed with that, there are three crucial issues to be tackled. First, some patient data is censored: we do not know the true survival times for all patients. Second, data is scarce, which led past research to treat different illness types as domains in a multi-task setup. Third, there is the need for adaptation to new or extremely rare illness types, where little or no labels are available. In contrast to previous multi-task setups, we want to investigate how to efficiently adapt to a new survival target domain from multiple survival source domains. For this, we introduce a new survival metric and the corresponding discrepancy measure between survival distributions. These allow us to define domain adaptation for survival analysis while incorporating censored data, which would otherwise have to be dropped. Our experiments on two cancer data sets reveal a superb performance on target domains, a better treatment recommendation, and a weight matrix with a plausible explanation.
comment: 37th AAAI Conference on Artificial Intelligence, 2023. Includes Appendix
♻ ☆ FedExP: Speeding Up Federated Averaging via Extrapolation ICLR 2023
Federated Averaging (FedAvg) remains the most popular algorithm for Federated Learning (FL) optimization due to its simple implementation, stateless nature, and privacy guarantees combined with secure aggregation. Recent work has sought to generalize the vanilla averaging in FedAvg to a generalized gradient descent step by treating client updates as pseudo-gradients and using a server step size. While the use of a server step size has been shown to provide performance improvement theoretically, the practical benefit of the server step size has not been seen in most existing works. In this work, we present FedExP, a method to adaptively determine the server step size in FL based on dynamically varying pseudo-gradients throughout the FL process. We begin by considering the overparameterized convex regime, where we reveal an interesting similarity between FedAvg and the Projection Onto Convex Sets (POCS) algorithm. We then show how FedExP can be motivated as a novel extension to the extrapolation mechanism that is used to speed up POCS. Our theoretical analysis later also discusses the implications of FedExP in underparameterized and non-convex settings. Experimental results show that FedExP consistently converges faster than FedAvg and competing baselines on a range of realistic FL datasets.
comment: Accepted to ICLR 2023. V2 fixes minor typos and cleans up proofs
♻ ☆ Automata Cascades: Expressivity and Sample Complexity AAAI 2023
Every automaton can be decomposed into a cascade of basic prime automata. This is the Prime Decomposition Theorem by Krohn and Rhodes. Guided by this theory, we propose automata cascades as a structured, modular, way to describe automata as complex systems made of many components, each implementing a specific functionality. Any automaton can serve as a component; using specific components allows for a fine-grained control of the expressivity of the resulting class of automata; using prime automata as components implies specific expressivity guarantees. Moreover, specifying automata as cascades allows for describing the sample complexity of automata in terms of their components. We show that the sample complexity is linear in the number of components and the maximum complexity of a single component, modulo logarithmic factors. This opens to the possibility of learning automata representing large dynamical systems consisting of many parts interacting with each other. It is in sharp contrast with the established understanding of the sample complexity of automata, described in terms of the overall number of states and input letters, which implies that it is only possible to learn automata where the number of states is linear in the amount of data available. Instead our results show that one can learn automata with a number of states that is exponential in the amount of data available.
comment: Full version with appendix of a paper with the same title that appears in the proceedings of AAAI 2023
♻ ☆ Evaluation of Interpretability Methods and Perturbation Artifacts in Deep Neural Networks
Despite excellent performance of deep neural networks (DNNs) in image classification, detection, and prediction, characterizing how DNNs make a given decision remains an open problem, resulting in a number of interpretability methods. Post-hoc interpretability methods primarily aim to quantify the importance of input features with respect to the class probabilities. However, due to the lack of ground truth and the existence of interpretability methods with diverse operating characteristics, evaluating these methods is a crucial challenge. A popular approach to evaluate interpretability methods is to perturb input features deemed important for a given prediction and observe the decrease in accuracy. However, perturbation itself may introduce artifacts, since perturbed images may be out-of-distribution (OOD). In this paper, we have conducted computational experiments to estimate the contribution of perturbation artifacts and developed a method to estimate the fidelity of interpretability methods. We demonstrate that, while perturbation artifacts indeed exist, we can minimize and characterize their impact on fidelity estimation by utilizing model accuracy curves from perturbing input features according to the Most Import First (MIF) and Least Import First (LIF) orders. Using the ResNet-50 trained on the ImageNet, we demonstrate the proposed fidelity estimation of four popular post-hoc interpretability methods.
comment: 26 pages, 9 figures
♻ ☆ Compositional Exemplars for In-context Learning
Large pretrained language models (LMs) have shown impressive In-Context Learning (ICL) ability, where the model learns to do an unseen task via a prompt consisting of input-output examples as the demonstration, without any parameter updates. The performance of ICL is highly dominated by the quality of the selected in-context examples. However, previous selection methods are mostly based on simple heuristics, leading to sub-optimal performance. In this work, we formulate in-context example selection as a subset selection problem. We propose CEIL (Compositional Exemplars for In-context Learning), which is instantiated by Determinantal Point Processes (DPPs) to model the interaction between the given input and in-context examples, and optimized through a carefully-designed contrastive learning objective to obtain preference from LMs. We validate CEIL on 12 classification and generation datasets from 7 distinct NLP tasks, including sentiment analysis, paraphrase detection, natural language inference, commonsense reasoning, open-domain question answering, code generation, and semantic parsing. Extensive experiments demonstrate not only the state-of-the-art performance but also the transferability and compositionality of CEIL, shedding new light on effective and efficient in-context learning. Our code is released at https://github.com/HKUNLP/icl-ceil.
♻ ☆ Lifting the Information Ratio: An Information-Theoretic Analysis of Thompson Sampling for Contextual Bandits
We study the Bayesian regret of the renowned Thompson Sampling algorithm in contextual bandits with binary losses and adversarially-selected contexts. We adapt the information-theoretic perspective of \cite{RvR16} to the contextual setting by considering a lifted version of the information ratio defined in terms of the unknown model parameter instead of the optimal action or optimal policy as done in previous works on the same setting. This allows us to bound the regret in terms of the entropy of the prior distribution through a remarkably simple proof, and with no structural assumptions on the likelihood or the prior. The extension to priors with infinite entropy only requires a Lipschitz assumption on the log-likelihood. An interesting special case is that of logistic bandits with $d$-dimensional parameters, $K$ actions, and Lipschitz logits, for which we provide a $\widetilde{O}(\sqrt{dKT})$ regret upper-bound that does not depend on the smallest slope of the sigmoid link function.
♻ ☆ Distribution-free Contextual Dynamic Pricing
Contextual dynamic pricing aims to set personalized prices based on sequential interactions with customers. At each time period, a customer who is interested in purchasing a product comes to the platform. The customer's valuation for the product is a linear function of contexts, including product and customer features, plus some random market noise. The seller does not observe the customer's true valuation, but instead needs to learn the valuation by leveraging contextual information and historical binary purchase feedbacks. Existing models typically assume full or partial knowledge of the random noise distribution. In this paper, we consider contextual dynamic pricing with unknown random noise in the valuation model. Our distribution-free pricing policy learns both the contextual function and the market noise simultaneously. A key ingredient of our method is a novel perturbed linear bandit framework, where a modified linear upper confidence bound algorithm is proposed to balance the exploration of market noise and the exploitation of the current knowledge for better pricing. We establish the regret upper bound and a matching lower bound of our policy in the perturbed linear bandit framework and prove a sub-linear regret bound in the considered pricing problem. Finally, we demonstrate the superior performance of our policy on simulations and a real-life auto-loan dataset.
comment: Accepted by Mathematics of Operations Research
♻ ☆ TPC: Transformation-Specific Smoothing for Point Cloud Models ICML 2022
Point cloud models with neural network architectures have achieved great success and have been widely used in safety-critical applications, such as Lidar-based recognition systems in autonomous vehicles. However, such models are shown vulnerable to adversarial attacks which aim to apply stealthy semantic transformations such as rotation and tapering to mislead model predictions. In this paper, we propose a transformation-specific smoothing framework TPC, which provides tight and scalable robustness guarantees for point cloud models against semantic transformation attacks. We first categorize common 3D transformations into three categories: additive (e.g., shearing), composable (e.g., rotation), and indirectly composable (e.g., tapering), and we present generic robustness certification strategies for all categories respectively. We then specify unique certification protocols for a range of specific semantic transformations and their compositions. Extensive experiments on several common 3D transformations show that TPC significantly outperforms the state of the art. For example, our framework boosts the certified accuracy against twisting transformation along z-axis (within 20$^\circ$) from 20.3$\%$ to 83.8$\%$. Codes and models are available at https://github.com/Qianhewu/Point-Cloud-Smoothing.
comment: Accepted as a conference paper at ICML 2022
♻ ☆ Which Factors are associated with Open Access Publishing? A Springer Nature Case Study
Open Access (OA) facilitates access to articles. But, authors or funders often must pay the publishing costs preventing authors who do not receive financial support from participating in OA publishing and citation advantage for OA articles. OA may exacerbate existing inequalities in the publication system rather than overcome them. To investigate this, we studied 522,411 articles published by Springer Nature. Employing correlation and regression analyses, we describe the relationship between authors affiliated with countries from different income levels, their choice of publishing model, and the citation impact of their papers. A machine learning classification method helped us to explore the importance of different features in predicting the publishing model. The results show that authors eligible for APC waivers publish more in gold-OA journals than others. In contrast, authors eligible for an APC discount have the lowest ratio of OA publications, leading to the assumption that this discount insufficiently motivates authors to publish in gold-OA journals. We found a strong correlation between the journal rank and the publishing model in gold-OA journals, whereas the OA option is mostly avoided in hybrid journals. Also, results show that the countries' income level, seniority, and experience with OA publications are the most predictive factors for OA publishing in hybrid journals.
♻ ☆ Is it enough to optimize CNN architectures on ImageNet?
Classification performance based on ImageNet is the de-facto standard metric for CNN development. In this work we challenge the notion that CNN architecture design solely based on ImageNet leads to generally effective convolutional neural network (CNN) architectures that perform well on a diverse set of datasets and application domains. To this end, we investigate and ultimately improve ImageNet as a basis for deriving such architectures. We conduct an extensive empirical study for which we train $500$ CNN architectures, sampled from the broad AnyNetX design space, on ImageNet as well as $8$ additional well known image classification benchmark datasets from a diverse array of application domains. We observe that the performances of the architectures are highly dataset dependent. Some datasets even exhibit a negative error correlation with ImageNet across all architectures. We show how to significantly increase these correlations by utilizing ImageNet subsets restricted to fewer classes. These contributions can have a profound impact on the way we design future CNN architectures and help alleviate the tilt we see currently in our community with respect to over-reliance on one dataset.
♻ ☆ User-friendly introduction to PAC-Bayes bounds
Aggregated predictors are obtained by making a set of basic predictors vote according to some weights, that is, to some probability distribution. Randomized predictors are obtained by sampling in a set of basic predictors, according to some prescribed probability distribution. Thus, aggregated and randomized predictors have in common that they are not defined by a minimization problem, but by a probability distribution on the set of predictors. In statistical learning theory, there is a set of tools designed to understand the generalization ability of such procedures: PAC-Bayesian or PAC-Bayes bounds. Since the original PAC-Bayes bounds of D. McAllester, these tools have been considerably improved in many directions (we will for example describe a simplified version of the localization technique of O. Catoni that was missed by the community, and later rediscovered as "mutual information bounds"). Very recently, PAC-Bayes bounds received a considerable attention: for example there was workshop on PAC-Bayes at NIPS 2017, "(Almost) 50 Shades of Bayesian Learning: PAC-Bayesian trends and insights", organized by B. Guedj, F. Bach and P. Germain. One of the reason of this recent success is the successful application of these bounds to neural networks by G. Dziugaite and D. Roy. An elementary introduction to PAC-Bayes theory is still missing. This is an attempt to provide such an introduction.
♻ ☆ A Reinforcement Learning Approach for Scheduling Problems With Improved Generalization Through Order Swapping
The scheduling of production resources (such as associating jobs to machines) plays a vital role for the manufacturing industry not only for saving energy but also for increasing the overall efficiency. Among the different job scheduling problems, the JSSP is addressed in this work. JSSP falls into the category of NP-hard COP, in which solving the problem through exhaustive search becomes unfeasible. Simple heuristics such as FIFO, LPT and metaheuristics such as Taboo search are often adopted to solve the problem by truncating the search space. The viability of the methods becomes inefficient for large problem sizes as it is either far from the optimum or time consuming. In recent years, the research towards using DRL to solve COP has gained interest and has shown promising results in terms of solution quality and computational efficiency. In this work, we provide an novel approach to solve the JSSP examining the objectives generalization and solution effectiveness using DRL. In particular, we employ the PPO algorithm that adopts the policy-gradient paradigm that is found to perform well in the constrained dispatching of jobs. We incorporated an OSM in the environment to achieve better generalized learning of the problem. The performance of the presented approach is analyzed in depth by using a set of available benchmark instances and comparing our results with the work of other groups.
♻ ☆ Adversarial Permutation Invariant Training for Universal Sound Separation ICASSP-23
Universal sound separation consists of separating mixes with arbitrary sounds of different types, and permutation invariant training (PIT) is used to train source agnostic models that do so. In this work, we complement PIT with adversarial losses but find it challenging with the standard formulation used in speech source separation. We overcome this challenge with a novel I-replacement context-based adversarial loss, and by training with multiple discriminators. Our experiments show that by simply improving the loss (keeping the same model and dataset) we obtain a non-negligible improvement of 1.4 dB SI-SNRi in the reverberant FUSS dataset. We also find adversarial PIT to be effective at reducing spectral holes, ubiquitous in mask-based separation models, which highlights the potential relevance of adversarial losses for source separation.
comment: Demo page: http://jordipons.me/apps/adversarialPIT/, Accepted at ICASSP-23
♻ ☆ Optimization of Convolutional Neural Network Using the Linearly Decreasing Weight Particle Swarm Optimization
Convolutional neural network (CNN) is one of the most frequently used deep learning techniques. Various forms of models have been proposed and im-proved for learning at CNN. When learning with CNN, it is necessary to determine the optimal hyperparameters. However, the number of hyperparameters is so large that it is difficult to do it manually, so much research has been done on automation. A method that uses metaheuristic algorithms is attracting attention in research on hyperparameter optimization. Metaheuristic algorithms are naturally inspired and include evolution strategies, genetic algorithms, antcolony optimization and particle swarm optimization. In particular, particle swarm optimization converges faster than genetic algorithms, and various models have been proposed. In this paper, we pro-pose CNN hyperparameter optimization with linearly decreasing weight particle swarm optimization (LDWPSO). In the experiment, the MNIST data set and CIFAR-10 data set, which are often used as benchmark data sets, are used. By opti-mizing CNN hyperparameters with LDWPSO, learning the MNIST and CIFAR-10 datasets, we compare the accuracy with a standard CNN based on LeNet-5. As a result, when using the MNIST dataset, the baseline CNN is 94.02% at the 5th epoch, compared to 98.95% for LDWPSO CNN, which improves accuracy. When using the CIFAR-10 dataset, the Baseline CNN is 28.07% at the 10th epoch, compared to 69.37% for the LDWPSO CNN, which greatly improves accuracy. This paper is presented at the 36th Annual Conference of the Japanese Society for Artificial In-telligence. The final version is available at the following URL: https://doi.org/10.11517/pjsai.JSAI2022.0_2S4IS2b03
comment: This paper is presented at the 36th Annual Conference of the Japanese Society for Artificial In-telligence
♻ ☆ MURANA: A Generic Framework for Stochastic Variance-Reduced Optimization
We propose a generic variance-reduced algorithm, which we call MUltiple RANdomized Algorithm (MURANA), for minimizing a sum of several smooth functions plus a regularizer, in a sequential or distributed manner. Our method is formulated with general stochastic operators, which allow us to model various strategies for reducing the computational complexity. For example, MURANA supports sparse activation of the gradients, and also reduction of the communication load via compression of the update vectors. This versatility allows MURANA to cover many existing randomization mechanisms within a unified framework, which also makes it possible to design new methods as special cases.
comment: 3rd Annual Conference on Mathematical and Scientific Machine Learning (MSML), Aug. 2022. PMLR 190:81-96
♻ ☆ VoxSRC 2022: The Fourth VoxCeleb Speaker Recognition Challenge
This paper summarises the findings from the VoxCeleb Speaker Recognition Challenge 2022 (VoxSRC-22), which was held in conjunction with INTERSPEECH 2022. The goal of this challenge was to evaluate how well state-of-the-art speaker recognition systems can diarise and recognise speakers from speech obtained "in the wild". The challenge consisted of: (i) the provision of publicly available speaker recognition and diarisation data from YouTube videos together with ground truth annotation and standardised evaluation software; and (ii) a public challenge and hybrid workshop held at INTERSPEECH 2022. We describe the four tracks of our challenge along with the baselines, methods, and results. We conclude with a discussion on the new domain-transfer focus of VoxSRC-22, and on the progression of the challenge from the previous three editions.
♻ ☆ Online certification of preference-based fairness for personalized recommender systems AAAI 2022
Recommender systems are facing scrutiny because of their growing impact on the opportunities we have access to. Current audits for fairness are limited to coarse-grained parity assessments at the level of sensitive groups. We propose to audit for envy-freeness, a more granular criterion aligned with individual preferences: every user should prefer their recommendations to those of other users. Since auditing for envy requires to estimate the preferences of users beyond their existing recommendations, we cast the audit as a new pure exploration problem in multi-armed bandits. We propose a sample-efficient algorithm with theoretical guarantees that it does not deteriorate user experience. We also study the trade-offs achieved on real-world recommendation datasets.
comment: AAAI 2022
♻ ☆ EF-BV: A Unified Theory of Error Feedback and Variance Reduction Mechanisms for Biased and Unbiased Compression in Distributed Optimization NeurIPS 2022
In distributed or federated optimization and learning, communication between the different computing units is often the bottleneck and gradient compression is widely used to reduce the number of bits sent within each communication round of iterative methods. There are two classes of compression operators and separate algorithms making use of them. In the case of unbiased random compressors with bounded variance (e.g., rand-k), the DIANA algorithm of Mishchenko et al. (2019), which implements a variance reduction technique for handling the variance introduced by compression, is the current state of the art. In the case of biased and contractive compressors (e.g., top-k), the EF21 algorithm of Richt\'arik et al. (2021), which instead implements an error-feedback mechanism, is the current state of the art. These two classes of compression schemes and algorithms are distinct, with different analyses and proof techniques. In this paper, we unify them into a single framework and propose a new algorithm, recovering DIANA and EF21 as particular cases. Our general approach works with a new, larger class of compressors, which has two parameters, the bias and the variance, and includes unbiased and biased compressors as particular cases. This allows us to inherit the best of the two worlds: like EF21 and unlike DIANA, biased compressors, like top-k, whose good performance in practice is recognized, can be used. And like DIANA and unlike EF21, independent randomness at the compressors allows to mitigate the effects of compression, with the convergence rate improving when the number of parallel workers is large. This is the first time that an algorithm with all these features is proposed. We prove its linear convergence under certain conditions. Our approach takes a step towards better understanding of two so-far distinct worlds of communication-efficient distributed learning.
comment: Conference NeurIPS 2022
♻ ☆ Leveraging Different Learning Styles for Improved Knowledge Distillation
Learning style refers to a type of training mechanism adopted by an individual to gain new knowledge. As suggested by the VARK model, humans have different learning preferences like visual, auditory, etc., for acquiring and effectively processing information. Inspired by this concept, our work explores the idea of mixed information sharing with model compression in the context of Knowledge Distillation (KD) and Mutual Learning (ML). Unlike conventional techniques that share the same type of knowledge with all networks, we propose to train individual networks with different forms of information to enhance the learning process. We formulate a combined KD and ML framework with one teacher and two student networks that share or exchange information in the form of predictions and feature maps. Our comprehensive experiments with benchmark classification and segmentation datasets demonstrate that with 15% compression, the ensemble performance of networks trained with diverse forms of knowledge outperforms the conventional techniques both quantitatively and qualitatively.
♻ ☆ Consistent Attack: Universal Adversarial Perturbation on Embodied Vision Navigation
Embodied agents in vision navigation coupled with deep neural networks have attracted increasing attention. However, deep neural networks have been shown vulnerable to malicious adversarial noises, which may potentially cause catastrophic failures in Embodied Vision Navigation. Among different adversarial noises, universal adversarial perturbations (UAP), i.e., a constant image-agnostic perturbation applied on every input frame of the agent, play a critical role in Embodied Vision Navigation since they are computation-efficient and application-practical during the attack. However, existing UAP methods ignore the system dynamics of Embodied Vision Navigation and might be sub-optimal. In order to extend UAP to the sequential decision setting, we formulate the disturbed environment under the universal noise $\delta$, as a $\delta$-disturbed Markov Decision Process ($\delta$-MDP). Based on the formulation, we analyze the properties of $\delta$-MDP and propose two novel Consistent Attack methods, named Reward UAP and Trajectory UAP, for attacking Embodied agents, which consider the dynamic of the MDP and calculate universal noises by estimating the disturbed distribution and the disturbed Q function. For various victim models, our Consistent Attack can cause a significant drop in their performance in the PointGoal task in Habitat with different datasets and different scenes. Extensive experimental results indicate that there exist serious potential risks for applying Embodied Vision Navigation methods to the real world.
♻ ☆ MetaFed: Federated Learning among Federations with Cyclic Knowledge Distillation for Personalized Healthcare IJCAI'22
Federated learning has attracted increasing attention to building models without accessing the raw user data, especially in healthcare. In real applications, different federations can seldom work together due to possible reasons such as data heterogeneity and distrust/inexistence of the central server. In this paper, we propose a novel framework called MetaFed to facilitate trustworthy FL between different federations. MetaFed obtains a personalized model for each federation without a central server via the proposed Cyclic Knowledge Distillation. Specifically, MetaFed treats each federation as a meta distribution and aggregates knowledge of each federation in a cyclic manner. The training is split into two parts: common knowledge accumulation and personalization. Comprehensive experiments on three benchmarks demonstrate that MetaFed without a server achieves better accuracy compared to state-of-the-art methods (e.g., 10%+ accuracy improvement compared to the baseline for PAMAP2) with fewer communication costs.
comment: IJCAI'22 FTL workshop innovation award; Extended version (12 pages) with more experiments and extensions; code at https://github.com/microsoft/PersonalizedFL
♻ ☆ Data-driven Modeling of Mach-Zehnder Interferometer-based Optical Matrix Multipliers
Photonic integrated circuits are facilitating the development of optical neural networks, which have the potential to be both faster and more energy efficient than their electronic counterparts since optical signals are especially well-suited for implementing matrix multiplications. However, accurate programming of photonic chips for optical matrix multiplication remains a difficult challenge. Here, we describe both simple analytical models and data-driven models for offline training of optical matrix multipliers. We train and evaluate the models using experimental data obtained from a fabricated chip featuring a Mach-Zehnder interferometer mesh implementing 3-by-3 matrix multiplication. The neural network-based models outperform the simple physics-based models in terms of prediction error. Furthermore, the neural network models are also able to predict the spectral variations in the matrix weights for up to 100 frequency channels covering the C-band. The use of neural network models for programming the chip for optical matrix multiplication yields increased performance on multiple machine learning tasks.
comment: 12 pages, 17 figures, submitted to Jorunal of Lightwave Technology
♻ ☆ A Federated Learning Benchmark for Drug-Target Interaction
Aggregating pharmaceutical data in the drug-target interaction (DTI) domain has the potential to deliver life-saving breakthroughs. It is, however, notoriously difficult due to regulatory constraints and commercial interests. This work proposes the application of federated learning, which we argue to be reconcilable with the industry's constraints, as it does not require sharing of any information that would reveal the entities' data or any other high-level summary of it. When used on a representative GraphDTA model and the KIBA dataset it achieves up to 15% improved performance relative to the best available non-privacy preserving alternative. Our extensive battery of experiments shows that, unlike in other domains, the non-IID data distribution in the DTI datasets does not deteriorate FL performance. Additionally, we identify a material trade-off between the benefits of adding new data, and the cost of adding more clients.
♻ ☆ NeuroExplainer: Fine-Grained Attention Decoding to Uncover Cortical Development Patterns of Preterm Infants
Deploying reliable deep learning techniques in interdisciplinary applications needs learned models to output accurate and ({even more importantly}) explainable predictions. Existing approaches typically explicate network outputs in a post-hoc fashion, under an implicit assumption that faithful explanations come from accurate predictions/classifications. We have an opposite claim that explanations boost (or even determine) classification. That is, end-to-end learning of explanation factors to augment discriminative representation extraction could be a more intuitive strategy to inversely assure fine-grained explainability, e.g., in those neuroimaging and neuroscience studies with high-dimensional data containing noisy, redundant, and task-irrelevant information. In this paper, we propose such an explainable geometric deep network dubbed as NeuroExplainer, with applications to uncover altered infant cortical development patterns associated with preterm birth. Given fundamental cortical attributes as network input, our NeuroExplainer adopts a hierarchical attention-decoding framework to learn fine-grained attentions and respective discriminative representations to accurately recognize preterm infants from term-born infants at term-equivalent age. NeuroExplainer learns the hierarchical attention-decoding modules under subject-level weak supervision coupled with targeted regularizers deduced from domain knowledge regarding brain development. These prior-guided constraints implicitly maximizes the explainability metrics (i.e., fidelity, sparsity, and stability) in network training, driving the learned network to output detailed explanations and accurate classifications. Experimental results on the public dHCP benchmark suggest that NeuroExplainer led to quantitatively reliable explanation results that are qualitatively consistent with representative neuroimaging studies.
comment: Some parts of the thesis are still being revised
♻ ☆ Synthetic Data Generator for Adaptive Interventions in Global Health
Artificial Intelligence and digital health have the potential to transform global health. However, having access to representative data to test and validate algorithms in realistic production environments is essential. We introduce HealthSyn, an open-source synthetic data generator of user behavior for testing reinforcement learning algorithms in the context of mobile health interventions. The generator utilizes Markov processes to generate diverse user actions, with individual user behavioral patterns that can change in reaction to personalized interventions (i.e., reminders, recommendations, and incentives). These actions are translated into actual logs using an ML-purposed data schema specific to the mobile health application functionality included with HealthKit, and open-source SDK. The logs can be fed to pipelines to obtain user metrics. The generated data, which is based on real-world behaviors and simulation techniques, can be used to develop, test, and evaluate, both ML algorithms in research and end-to-end operational RL-based intervention delivery frameworks.
♻ ☆ Fixed-budget online adaptive learning for physics-informed neural networks. Towards parameterized problem inference
Physics-Informed Neural Networks (PINNs) have gained much attention in various fields of engineering thanks to their capability of incorporating physical laws into the models. PINNs integrate the physical constraints by minimizing the partial differential equations (PDEs) residuals on a set of collocation points. The distribution of these collocation points appears to have a huge impact on the performance of PINNs and the assessment of the sampling methods for these points is still an active topic. In this paper, we propose a Fixed-Budget Online Adaptive Learning (FBOAL) method, which decomposes the domain into sub-domains, for training collocation points based on local maxima and local minima of the PDEs residuals. The effectiveness of FBOAL is demonstrated for non-parameterized and parameterized problems. The comparison with other adaptive sampling methods is also illustrated. The numerical results demonstrate important gains in terms of the accuracy and computational cost of PINNs with FBOAL over the classical PINNs with non-adaptive collocation points. We also apply FBOAL in a complex industrial application involving coupling between mechanical and thermal fields. We show that FBOAL is able to identify the high-gradient locations and even give better predictions for some physical fields than the classical PINNs with collocation points sampled on a pre-adapted finite element mesh built thanks to numerical expert knowledge. From the present study, it is expected that the use of FBOAL will help to improve the conventional numerical solver in the construction of the mesh.
comment: 22 pages, 30 figures, 3 tables
♻ ☆ Sampling-free Inference for Ab-Initio Potential Energy Surface Networks ICLR 2023
Recently, it has been shown that neural networks not only approximate the ground-state wave functions of a single molecular system well but can also generalize to multiple geometries. While such generalization significantly speeds up training, each energy evaluation still requires Monte Carlo integration which limits the evaluation to a few geometries. In this work, we address the inference shortcomings by proposing the Potential learning from ab-initio Networks (PlaNet) framework, in which we simultaneously train a surrogate model in addition to the neural wave function. At inference time, the surrogate avoids expensive Monte-Carlo integration by directly estimating the energy, accelerating the process from hours to milliseconds. In this way, we can accurately model high-resolution multi-dimensional energy surfaces for larger systems that previously were unobtainable via neural wave functions. Finally, we explore an additional inductive bias by introducing physically-motivated restricted neural wave function models. We implement such a function with several additional improvements in the new PESNet++ model. In our experimental evaluation, PlaNet accelerates inference by 7 orders of magnitude for larger molecules like ethanol while preserving accuracy. Compared to previous energy surface networks, PESNet++ reduces energy errors by up to 74%.
comment: Published as a conference paper at ICLR 2023
♻ ☆ A Categorical Archive of ChatGPT Failures
Large language models have been demonstrated to be valuable in different fields. ChatGPT, developed by OpenAI, has been trained using massive amounts of data and simulates human conversation by comprehending context and generating appropriate responses. It has garnered significant attention due to its ability to effectively answer a broad range of human inquiries, with fluent and comprehensive answers surpassing prior public chatbots in both security and usefulness. However, a comprehensive analysis of ChatGPT's failures is lacking, which is the focus of this study. Eleven categories of failures, including reasoning, factual errors, math, coding, and bias, are presented and discussed. The risks, limitations, and societal implications of ChatGPT are also highlighted. The goal of this study is to assist researchers and developers in enhancing future language models and chatbots.
♻ ☆ Deep Double Descent via Smooth Interpolation
The ability of overparameterized deep networks to interpolate noisy data, while at the same time showing good generalization performance, has been recently characterized in terms of the double descent curve for the test error. Common intuition from polynomial regression suggests that overparameterized networks are able to sharply interpolate noisy data, without considerably deviating from the ground-truth signal, thus preserving their generalization ability. At present, a precise characterization of the relationship between interpolation and generalization for deep networks is missing. In this work, we quantify sharpness of fit of the training data interpolated by neural network functions, by studying the loss landscape w.r.t.\ to the input variable locally to each training point, over volumes around cleanly- and noisily-labelled training samples, as we systematically increase the number of model parameters and training epochs. Our findings show that loss sharpness in the input space follows both model- and epoch-wise double descent, with worse peaks observed around noisy labels. While small interpolating models sharply fit both clean and noisy data, large interpolating models express a smooth loss landscape, where noisy targets are predicted over large volumes around training data points, in contrast to existing intuition.
♻ ☆ Learning Discretized Neural Networks under Ricci Flow
In this paper, we consider Discretized Neural Networks (DNNs) consisting of low-precision weights and activations, which suffer from either infinite or zero gradients caused by the non-differentiable discrete function in the training process. In this case, most training-based DNNs use the standard Straight-Through Estimator (STE) to approximate the gradient w.r.t. discrete values. However, the STE will cause the problem of gradient mismatch, which implies that the approximated gradient is with perturbations. We propose an analysis that this mismatch can be viewed as a metric perturbation in a Riemannian manifold through the lens of duality theory. To address this problem, based on the information geometry, we construct the Linearly Nearly Euclidean (LNE) manifold for DNNs as a background to deal with perturbations. By introducing a partial differential equation on metrics, the Ricci flow, we prove the dynamical stability and convergence of the LNE metric with the $L^2$-norm perturbation. And unlike the previous perturbation theory which gives the rate of convergence is the fractional powers, we yield the metric perturbation under the Ricci flow can be exponentially decayed in the LNE manifold. The experimental results on various datasets demonstrate that our method achieves better and more stable performance for DNNs than other representative training-based methods.
♻ ☆ Quantifying Memorization Across Neural Language Models
Large language models (LMs) have been shown to memorize parts of their training data, and when prompted appropriately, they will emit the memorized training data verbatim. This is undesirable because memorization violates privacy (exposing user data), degrades utility (repeated easy-to-memorize text is often low quality), and hurts fairness (some texts are memorized over others). We describe three log-linear relationships that quantify the degree to which LMs emit memorized training data. Memorization significantly grows as we increase (1) the capacity of a model, (2) the number of times an example has been duplicated, and (3) the number of tokens of context used to prompt the model. Surprisingly, we find the situation becomes more complicated when generalizing these results across model families. On the whole, we find that memorization in LMs is more prevalent than previously believed and will likely get worse as models continues to scale, at least without active mitigations.
♻ ☆ (Certified!!) Adversarial Robustness for Free!
In this paper we show how to achieve state-of-the-art certified adversarial robustness to 2-norm bounded perturbations by relying exclusively on off-the-shelf pretrained models. To do so, we instantiate the denoised smoothing approach of Salman et al. 2020 by combining a pretrained denoising diffusion probabilistic model and a standard high-accuracy classifier. This allows us to certify 71% accuracy on ImageNet under adversarial perturbations constrained to be within an 2-norm of 0.5, an improvement of 14 percentage points over the prior certified SoTA using any approach, or an improvement of 30 percentage points over denoised smoothing. We obtain these results using only pretrained diffusion models and image classifiers, without requiring any fine tuning or retraining of model parameters.
♻ ☆ Coverage-centric Coreset Selection for High Pruning Rates ICLR
One-shot coreset selection aims to select a representative subset of the training data, given a pruning rate, that can later be used to train future models while retaining high accuracy. State-of-the-art coreset selection methods pick the highest importance examples based on an importance metric and are found to perform well at low pruning rates. However, at high pruning rates, they suffer from a catastrophic accuracy drop, performing worse than even random sampling. This paper explores the reasons behind this accuracy drop both theoretically and empirically. We first propose a novel metric to measure the coverage of a dataset on a specific distribution by extending the classical geometric set cover problem to a distribution cover problem. This metric helps explain why coresets selected by SOTA methods at high pruning rates perform poorly compared to random sampling because of worse data coverage. We then propose a novel one-shot coreset selection method, Coverage-centric Coreset Selection (CCS), that jointly considers overall data coverage upon a distribution as well as the importance of each example. We evaluate CCS on five datasets and show that, at high pruning rates (e.g., 90%), it achieves significantly better accuracy than previous SOTA methods (e.g., at least 19.56% higher on CIFAR10) as well as random selection (e.g., 7.04% higher on CIFAR10) and comparable accuracy at low pruning rates. We make our code publicly available at https://github.com/haizhongzheng/Coverage-centric-coreset-selection.
comment: International Conference on Learning Representations (ICLR) 2023
♻ ☆ The Point to Which Soft Actor-Critic Converges
Soft actor-critic is a successful successor over soft Q-learning. While lived under maximum entropy framework, their relationship is still unclear. In this paper, we prove that in the limit they converge to the same solution. This is appealing since it translates the optimization from an arduous to an easier way. The same justification can also be applied to other regularizers such as KL divergence.
♻ ☆ Switchable Representation Learning Framework with Self-compatibility
Real-world visual search systems involve deployments on multiple platforms with different computing and storage resources. Deploying a unified model that suits the minimal-constrain platforms leads to limited accuracy. It is expected to deploy models with different capacities adapting to the resource constraints, which requires features extracted by these models to be aligned in the metric space. The method to achieve feature alignments is called ``compatible learning''. Existing research mainly focuses on the one-to-one compatible paradigm, which is limited in learning compatibility among multiple models. We propose a \textbf{S}witchable representation learning Framework with Self-Compatibility (SFSC). SFSC generates a series of compatible sub-models with different capacities through one training process. The optimization of sub-models faces gradients conflict, and we mitigate this problem from the perspective of the magnitude and direction. We adjust the priorities of sub-models dynamically through uncertainty estimation to co-optimize sub-models properly. Besides, the gradients with conflicting directions are projected to avoid mutual interference. SFSC achieves state-of-the-art performance on the evaluated datasets.
♻ ☆ Understanding and Unifying Fourteen Attribution Methods with Taylor Interactions
Various attribution methods have been developed to explain deep neural networks (DNNs) by inferring the attribution/importance/contribution score of each input variable to the final output. However, existing attribution methods are often built upon different heuristics. There remains a lack of a unified theoretical understanding of why these methods are effective and how they are related. To this end, for the first time, we formulate core mechanisms of fourteen attribution methods, which were designed on different heuristics, into the same mathematical system, i.e., the system of Taylor interactions. Specifically, we prove that attribution scores estimated by fourteen attribution methods can all be reformulated as the weighted sum of two types of effects, i.e., independent effects of each individual input variable and interaction effects between input variables. The essential difference among the fourteen attribution methods mainly lies in the weights of allocating different effects. Based on the above findings, we propose three principles for a fair allocation of effects to evaluate the faithfulness of the fourteen attribution methods.
♻ ☆ Need for Objective Task-based Evaluation of Deep Learning-Based Denoising Methods: A Study in the Context of Myocardial Perfusion SPECT
Artificial intelligence-based methods have generated substantial interest in nuclear medicine. An area of significant interest has been using deep-learning (DL)-based approaches for denoising images acquired with lower doses, shorter acquisition times, or both. Objective evaluation of these approaches is essential for clinical application. DL-based approaches for denoising nuclear-medicine images have typically been evaluated using fidelity-based figures of merit (FoMs) such as RMSE and SSIM. However, these images are acquired for clinical tasks and thus should be evaluated based on their performance in these tasks. Our objectives were to (1) investigate whether evaluation with these FoMs is consistent with objective clinical-task-based evaluation; (2) provide a theoretical analysis for determining the impact of denoising on signal-detection tasks; (3) demonstrate the utility of virtual clinical trials (VCTs) to evaluate DL-based methods. A VCT to evaluate a DL-based method for denoising myocardial perfusion SPECT (MPS) images was conducted. The impact of DL-based denoising was evaluated using fidelity-based FoMs and AUC, which quantified performance on detecting perfusion defects in MPS images as obtained using a model observer with anthropomorphic channels. Based on fidelity-based FoMs, denoising using the considered DL-based method led to significantly superior performance. However, based on ROC analysis, denoising did not improve, and in fact, often degraded detection-task performance. The results motivate the need for objective task-based evaluation of DL-based denoising approaches. Further, this study shows how VCTs provide a mechanism to conduct such evaluations using VCTs. Finally, our theoretical treatment reveals insights into the reasons for the limited performance of the denoising approach.
♻ ☆ Data-Centric AI: Deep Generative Differentiable Feature Selection via Discrete Subsetting as Continuous Embedding Space Optimization
Feature Selection (FS), such as filter, wrapper, and embedded methods, aims to find the optimal feature subset for a given downstream task. However, in many real-world practices, 1) the criteria of FS vary across domains; 2) FS is brittle when data is a high-dimensional and small sample size. Can selected feature subsets be more generalized, accurate, and input dimensionality agnostic? We generalize this problem into a deep differentiable feature selection task and propose a new perspective: discrete feature subsetting as continuous embedding space optimization. We develop a generic and principled framework including a deep feature subset encoder, accuracy evaluator, decoder, and gradient ascent optimizer. This framework implements four steps: 1) features-accuracy training data preparation; 2) deep feature subset embedding; 3) gradient-optimized search; 4) feature subset reconstruction. We develop new technical insights: reinforcement as a training data generator, ensembles of diverse peer and exploratory feature selector knowledge for generalization, an effective embedding from feature subsets to continuous space along with joint optimizing reconstruction and accuracy losses to select accurate features. Experimental results demonstrate the effectiveness of the proposed method.
comment: 8 pages
♻ ☆ T-Cal: An optimal test for the calibration of predictive models
The prediction accuracy of machine learning methods is steadily increasing, but the calibration of their uncertainty predictions poses a significant challenge. Numerous works focus on obtaining well-calibrated predictive models, but less is known about reliably assessing model calibration. This limits our ability to know when algorithms for improving calibration have a real effect, and when their improvements are merely artifacts due to random noise in finite datasets. In this work, we consider detecting mis-calibration of predictive models using a finite validation dataset as a hypothesis testing problem. The null hypothesis is that the predictive model is calibrated, while the alternative hypothesis is that the deviation from calibration is sufficiently large. We find that detecting mis-calibration is only possible when the conditional probabilities of the classes are sufficiently smooth functions of the predictions. When the conditional class probabilities are H\"older continuous, we propose T-Cal, a minimax optimal test for calibration based on a debiased plug-in estimator of the $\ell_2$-Expected Calibration Error (ECE). We further propose Adaptive T-Cal, a version that is adaptive to unknown smoothness. We verify our theoretical findings with a broad range of experiments, including with several popular deep neural net architectures and several standard post-hoc calibration methods. T-Cal is a practical general-purpose tool, which -- combined with classical tests for discrete-valued predictors -- can be used to test the calibration of virtually any probabilistic classification method.
comment: The implementation of T-Cal is available at https://github.com/dh7401/T-Cal
♻ ☆ Augmentation-Free Graph Contrastive Learning of Invariant-Discriminative Representations
The pretasks are mainly built on mutual information estimation, which requires data augmentation to construct positive samples with similar semantics to learn invariant signals and negative samples with dissimilar semantics in order to empower representation discriminability. However, an appropriate data augmentation configuration depends heavily on lots of empirical trials such as choosing the compositions of data augmentation techniques and the corresponding hyperparameter settings. We propose an augmentation-free graph contrastive learning method, invariant-discriminative graph contrastive learning (iGCL), that does not intrinsically require negative samples. iGCL designs the invariant-discriminative loss (ID loss) to learn invariant and discriminative representations. On the one hand, ID loss learns invariant signals by directly minimizing the mean square error between the target samples and positive samples in the representation space. On the other hand, ID loss ensures that the representations are discriminative by an orthonormal constraint forcing the different dimensions of representations to be independent of each other. This prevents representations from collapsing to a point or subspace. Our theoretical analysis explains the effectiveness of ID loss from the perspectives of the redundancy reduction criterion, canonical correlation analysis, and information bottleneck principle. The experimental results demonstrate that iGCL outperforms all baselines on 5 node classification benchmark datasets. iGCL also shows superior performance for different label ratios and is capable of resisting graph attacks, which indicates that iGCL has excellent generalization and robustness. The source code is available at https://github.com/lehaifeng/T-GCN/tree/master/iGCL.
comment: 11 pages 8 figs
♻ ☆ Backdoor Defense via Suppressing Model Shortcuts ICASSP 2023
Recent studies have demonstrated that deep neural networks (DNNs) are vulnerable to backdoor attacks during the training process. Specifically, the adversaries intend to embed hidden backdoors in DNNs so that malicious model predictions can be activated through pre-defined trigger patterns. In this paper, we explore the backdoor mechanism from the angle of the model structure. We select the skip connection for discussions, inspired by the understanding that it helps the learning of model `shortcuts' where backdoor triggers are usually easier to be learned. Specifically, we demonstrate that the attack success rate (ASR) decreases significantly when reducing the outputs of some key skip connections. Based on this observation, we design a simple yet effective backdoor removal method by suppressing the skip connections in critical layers selected by our method. We also implement fine-tuning on these layers to recover high benign accuracy and to further reduce ASR. Extensive experiments on benchmark datasets verify the effectiveness of our method.
comment: This paper is accepted by ICASSP 2023. 5 pages
♻ ☆ BATT: Backdoor Attack with Transformation-based Triggers ICASSP 2023
Deep neural networks (DNNs) are vulnerable to backdoor attacks. The backdoor adversaries intend to maliciously control the predictions of attacked DNNs by injecting hidden backdoors that can be activated by adversary-specified trigger patterns during the training process. One recent research revealed that most of the existing attacks failed in the real physical world since the trigger contained in the digitized test samples may be different from that of the one used for training. Accordingly, users can adopt spatial transformations as the image pre-processing to deactivate hidden backdoors. In this paper, we explore the previous findings from another side. We exploit classical spatial transformations (i.e. rotation and translation) with the specific parameter as trigger patterns to design a simple yet effective poisoning-based backdoor attack. For example, only images rotated to a particular angle can activate the embedded backdoor of attacked DNNs. Extensive experiments are conducted, verifying the effectiveness of our attack under both digital and physical settings and its resistance to existing backdoor defenses.
comment: This paper is accepted by ICASSP 2023. 5 pages
♻ ☆ Untargeted Backdoor Attack against Object Detection ICASSP 2023
Recent studies revealed that deep neural networks (DNNs) are exposed to backdoor threats when training with third-party resources (such as training samples or backbones). The backdoored model has promising performance in predicting benign samples, whereas its predictions can be maliciously manipulated by adversaries based on activating its backdoors with pre-defined trigger patterns. Currently, most of the existing backdoor attacks were conducted on the image classification under the targeted manner. In this paper, we reveal that these threats could also happen in object detection, posing threatening risks to many mission-critical applications ($e.g.$, pedestrian detection and intelligent surveillance systems). Specifically, we design a simple yet effective poison-only backdoor attack in an untargeted manner, based on task characteristics. We show that, once the backdoor is embedded into the target model by our attack, it can trick the model to lose detection of any object stamped with our trigger patterns. We conduct extensive experiments on the benchmark dataset, showing its effectiveness in both digital and physical-world settings and its resistance to potential defenses.
comment: This paper is accepted by ICASSP 2023. 5 pages
♻ ☆ Smoothness Analysis of Adversarial Training
Deep neural networks are vulnerable to adversarial attacks. Recent studies about adversarial robustness focus on the loss landscape in the parameter space since it is related to optimization and generalization performance. These studies conclude that the difficulty of adversarial training is caused by the non-smoothness of the loss function: i.e., its gradient is not Lipschitz continuous. However, this analysis ignores the dependence of adversarial attacks on model parameters. Since adversarial attacks are optimized for models, they should depend on the parameters. Considering this dependence, we analyze the smoothness of the loss function of adversarial training using the optimal attacks for the model parameter in more detail. We reveal that the constraint of adversarial attacks is one cause of the non-smoothness and that the smoothness depends on the types of the constraints. Specifically, the $L_\infty$ constraint can cause non-smoothness more than the $L_2$ constraint. Moreover, our analysis implies that if we flatten the loss function with respect to input data, the Lipschitz constant of the gradient of adversarial loss tends to increase. To address the non-smoothness, we show that EntropySGD smoothens the non-smooth loss and improves the performance of adversarial training.
comment: The latest version of this article is published in IEEE Transactions on Neural Networks and Learning Systems (DOI: 10.1109/TNNLS.2023.3244172). 22 pages, 7 figures
♻ ☆ The Optimal Choice of Hypothesis Is the Weakest, Not the Shortest
If $A$ and $B$ are sets such that $A \subset B$, generalisation may be understood as the inference from $A$ of a hypothesis sufficient to construct $B$. One might infer any number of hypotheses from $A$, yet only some of those may generalise to $B$. How can one know which are likely to generalise? One strategy is to choose the shortest, equating the ability to compress information with the ability to generalise (a proxy for intelligence). We examine this in the context of a mathematical formalism of enactive cognition. We show that compression is neither necessary nor sufficient to maximise performance (measured in terms of the probability of a hypothesis generalising). We formulate a proxy unrelated to length or simplicity, called weakness. We show that if tasks are uniformly distributed, then there is no choice of proxy that performs at least as well as weakness maximisation in all tasks while performing strictly better in at least one. In other words, weakness is the pareto optimal choice of proxy. In experiments comparing maximum weakness and minimum description length in the context of binary arithmetic, the former generalised at between $1.1$ and $5$ times the rate of the latter. We argue this demonstrates that weakness is a far better proxy, and explains why Deepmind's Apperception Engine is able to generalise effectively.
♻ ☆ Denoising diffusion models for out-of-distribution detection
Out-of-distribution detection is crucial to the safe deployment of machine learning systems. Currently, unsupervised out-of-distribution detection is dominated by generative-based approaches that make use of estimates of the likelihood or other measurements from a generative model. Reconstruction-based methods offer an alternative approach, in which a measure of reconstruction error is used to determine if a sample is out-of-distribution. However, reconstruction-based approaches are less favoured, as they require careful tuning of the model's information bottleneck - such as the size of the latent dimension - to produce good results. In this work, we exploit the view of denoising diffusion probabilistic models (DDPM) as denoising autoencoders where the bottleneck is controlled externally, by means of the amount of noise applied. We propose to use DDPMs to reconstruct an input that has been noised to a range of noise levels, and use the resulting multi-dimensional reconstruction error to classify out-of-distribution inputs. We validate our approach both on standard computer-vision datasets and on higher dimension medical datasets. Our approach outperforms not only reconstruction-based methods, but also state-of-the-art generative-based approaches.
Hyena Hierarchy: Towards Larger Convolutional Language Models
Recent advances in deep learning have relied heavily on the use of large Transformers due to their ability to learn at scale. However, the core building block of Transformers, the attention operator, exhibits quadratic cost in sequence length, limiting the amount of context accessible. Existing subquadratic methods based on low-rank and sparse approximations need to be combined with dense attention layers to match Transformers, indicating a gap in capability. In this work, we propose Hyena, a subquadratic drop-in replacement for attention constructed by interleaving implicitly parametrized long convolutions and data-controlled gating. In recall and reasoning tasks on sequences of thousands to hundreds of thousands of tokens, Hyena improves accuracy by more than 50 points over operators relying on state-spaces and other implicit and explicit methods, matching attention-based models. We set a new state-of-the-art for dense-attention-free architectures on language modeling in standard datasets (WikiText103 and The Pile), reaching Transformer quality with a 20% reduction in training compute required at sequence length 2K. Hyena operators are twice as fast as highly optimized attention at sequence length 8K, and 100x faster at sequence length 64K.
comment: Additional results (PG-19, LAMBADA)
♻ ☆ Learning Explicit Credit Assignment for Cooperative Multi-Agent Reinforcement Learning via Polarization Policy Gradient
Cooperative multi-agent policy gradient (MAPG) algorithms have recently attracted wide attention and are regarded as a general scheme for the multi-agent system. Credit assignment plays an important role in MAPG and can induce cooperation among multiple agents. However, most MAPG algorithms cannot achieve good credit assignment because of the game-theoretic pathology known as \textit{centralized-decentralized mismatch}. To address this issue, this paper presents a novel method, \textit{\underline{M}ulti-\underline{A}gent \underline{P}olarization \underline{P}olicy \underline{G}radient} (MAPPG). MAPPG takes a simple but efficient polarization function to transform the optimal consistency of joint and individual actions into easily realized constraints, thus enabling efficient credit assignment in MAPG. Theoretically, we prove that individual policies of MAPPG can converge to the global optimum. Empirically, we evaluate MAPPG on the well-known matrix game and differential game, and verify that MAPPG can converge to the global optimum for both discrete and continuous action spaces. We also evaluate MAPPG on a set of StarCraft II micromanagement tasks and demonstrate that MAPPG outperforms the state-of-the-art MAPG algorithms.
♻ ☆ An Unpooling Layer for Graph Generation
We propose a novel and trainable graph unpooling layer for effective graph generation. Given a graph with features, the unpooling layer enlarges this graph and learns its desired new structure and features. Since this unpooling layer is trainable, it can be applied to graph generation either in the decoder of a variational autoencoder or in the generator of a generative adversarial network (GAN). We prove that the unpooled graph remains connected and any connected graph can be sequentially unpooled from a 3-nodes graph. We apply the unpooling layer within the GAN generator. Since the most studied instance of graph generation is molecular generation, we test our ideas in this context. Using the QM9 and ZINC datasets, we demonstrate the improvement obtained by using the unpooling layer instead of an adjacency-matrix-based approach.
♻ ☆ DeepStruct: Pretraining of Language Models for Structure Prediction ACL 2022
We introduce a method for improving the structural understanding abilities of language models. Unlike previous approaches that finetune the models with task-specific augmentation, we pretrain language models on a collection of task-agnostic corpora to generate structures from text. Our structure pretraining enables zero-shot transfer of the learned knowledge that models have about the structure tasks. We study the performance of this approach on 28 datasets, spanning 10 structure prediction tasks including open information extraction, joint entity and relation extraction, named entity recognition, relation classification, semantic role labeling, event extraction, coreference resolution, factual probe, intent detection, and dialogue state tracking. We further enhance the pretraining with the task-specific training sets. We show that a 10B parameter language model transfers non-trivially to most tasks and obtains state-of-the-art performance on 21 of 28 datasets that we evaluate.
comment: ACL 2022
♻ ☆ Benford's law: what does it say on adversarial images?
Convolutional neural networks (CNNs) are fragile to small perturbations in the input images. These networks are thus prone to malicious attacks that perturb the inputs to force a misclassification. Such slightly manipulated images aimed at deceiving the classifier are known as adversarial images. In this work, we investigate statistical differences between natural images and adversarial ones. More precisely, we show that employing a proper image transformation and for a class of adversarial attacks, the distribution of the leading digit of the pixels in adversarial images deviates from Benford's law. The stronger the attack, the more distant the resulting distribution is from Benford's law. Our analysis provides a detailed investigation of this new approach that can serve as a basis for alternative adversarial example detection methods that do not need to modify the original CNN classifier neither work on the raw high-dimensional pixels as features to defend against attacks.
♻ ☆ ReLOAD: Reinforcement Learning with Optimistic Ascent-Descent for Last-Iterate Convergence in Constrained MDPs
In recent years, Reinforcement Learning (RL) has been applied to real-world problems with increasing success. Such applications often require to put constraints on the agent's behavior. Existing algorithms for constrained RL (CRL) rely on gradient descent-ascent, but this approach comes with a caveat. While these algorithms are guaranteed to converge on average, they do not guarantee last-iterate convergence, i.e., the current policy of the agent may never converge to the optimal solution. In practice, it is often observed that the policy alternates between satisfying the constraints and maximizing the reward, rarely accomplishing both objectives simultaneously. Here, we address this problem by introducing Reinforcement Learning with Optimistic Ascent-Descent (ReLOAD), a principled CRL method with guaranteed last-iterate convergence. We demonstrate its empirical effectiveness on a wide variety of CRL problems including discrete MDPs and continuous control. In the process we establish a benchmark of challenging CRL problems.
♻ ☆ Time-Series Pattern Recognition in Smart Manufacturing Systems: A Literature Review and Ontology
Since the inception of Industry 4.0 in 2012, emerging technologies have enabled the acquisition of vast amounts of data from diverse sources such as machine tools, robust and affordable sensor systems with advanced information models, and other sources within Smart Manufacturing Systems (SMS). As a result, the amount of data that is available in manufacturing settings has exploded, allowing data-hungry tools such as Artificial Intelligence (AI) and Machine Learning (ML) to be leveraged. Time-series analytics has been successfully applied in a variety of industries, and that success is now being migrated to pattern recognition applications in manufacturing to support higher quality products, zero defect manufacturing, and improved customer satisfaction. However, the diverse landscape of manufacturing presents a challenge for successfully solving problems in industry using time-series pattern recognition. The resulting research gap of understanding and applying the subject matter of time-series pattern recognition in manufacturing is a major limiting factor for adoption in industry. The purpose of this paper is to provide a structured perspective of the current state of time-series pattern recognition in manufacturing with a problem-solving focus. By using an ontology to classify and define concepts, how they are structured, their properties, the relationships between them, and considerations when applying them, this paper aims to provide practical and actionable guidelines for application and recommendations for advancing time-series analytics.
comment: 40 pages, 35 figures, 21 tables, submitted to Elsevier
♻ ☆ Temporal Dependencies in Feature Importance for Time Series Predictions ICLR'23
Time series data introduces two key challenges for explainability methods: firstly, observations of the same feature over subsequent time steps are not independent, and secondly, the same feature can have varying importance to model predictions over time. In this paper, we propose Windowed Feature Importance in Time (WinIT), a feature removal based explainability approach to address these issues. Unlike existing feature removal explanation methods, WinIT explicitly accounts for the temporal dependence between different observations of the same feature in the construction of its importance score. Furthermore, WinIT captures the varying importance of a feature over time, by summarizing its importance over a window of past time steps. We conduct an extensive empirical study on synthetic and real-world data, compare against a wide range of leading explainability methods, and explore the impact of various evaluation strategies. Our results show that WinIT achieves significant gains over existing methods, with more consistent performance across different evaluation metrics. The code for our work is publicly available at \url{https://github.com/layer6ai-labs/WinIT}.
comment: International Conference on Learning Representations 2023 (ICLR'23)
♻ ☆ How Much Space Has Been Explored? Measuring the Chemical Space Covered by Databases and Machine-Generated Molecules ICLR 2023
Forming a molecular candidate set that contains a wide range of potentially effective compounds is crucial to the success of drug discovery. While most databases and machine-learning-based generation models aim to optimize particular chemical properties, there is limited literature on how to properly measure the coverage of the chemical space by those candidates included or generated. This problem is challenging due to the lack of formal criteria to select good measures of the chemical space. In this paper, we propose a novel evaluation framework for measures of the chemical space based on two analyses: an axiomatic analysis with three intuitive axioms that a good measure should obey, and an empirical analysis on the correlation between a measure and a proxy gold standard. Using this framework, we are able to identify #Circles, a new measure of chemical space coverage, which is superior to existing measures both analytically and empirically. We further evaluate how well the existing databases and generation models cover the chemical space in terms of #Circles. The results suggest that many generation models fail to explore a larger space over existing databases, which leads to new opportunities for improving generation models by encouraging exploration.
comment: ICLR 2023
♻ ☆ Bridging the Gap to Real-World Object-Centric Learning ICLR 2023
Humans naturally decompose their environment into entities at the appropriate level of abstraction to act in the world. Allowing machine learning algorithms to derive this decomposition in an unsupervised way has become an important line of research. However, current methods are restricted to simulated data or require additional information in the form of motion or depth in order to successfully discover objects. In this work, we overcome this limitation by showing that reconstructing features from models trained in a self-supervised manner is a sufficient training signal for object-centric representations to arise in a fully unsupervised way. Our approach, DINOSAUR, significantly out-performs existing image-based object-centric learning models on simulated data and is the first unsupervised object-centric model that scales to real-world datasets such as COCO and PASCAL VOC. DINOSAUR is conceptually simple and shows competitive performance compared to more involved pipelines from the computer vision literature.
comment: ICLR 2023 camera-ready version
♻ ☆ Nearly Minimax Optimal Reinforcement Learning for Linear Markov Decision Processes
We study reinforcement learning (RL) with linear function approximation. For episodic time-inhomogeneous linear Markov decision processes (linear MDPs) whose transition dynamic can be parameterized as a linear function of a given feature mapping, we propose the first computationally efficient algorithm that achieves the nearly minimax optimal regret $\tilde O(d\sqrt{H^3K})$, where $d$ is the dimension of the feature mapping, $H$ is the planning horizon, and $K$ is the number of episodes. Our algorithm is based on a weighted linear regression scheme with a carefully designed weight, which depends on a new variance estimator that (1) directly estimates the variance of the \emph{optimal} value function, (2) monotonically decreases with respect to the number of episodes to ensure a better estimation accuracy, and (3) uses a rare-switching policy to update the value function estimator to control the complexity of the estimated value function class. Our work provides a complete answer to optimal RL with linear MDPs, and the developed algorithm and theoretical tools may be of independent interest.
comment: 44 pages, 1 table
♻ ☆ Scale up with Order: Finding Good Data Permutations for Distributed Training
Gradient Balancing (GraB) is a recently proposed technique that finds provably better data permutations when training models with multiple epochs over a finite dataset. It converges at a faster rate than the widely adopted Random Reshuffling, by minimizing the discrepancy of the gradients on adjacently selected examples. However, GraB only operates under critical assumptions such as small batch sizes and centralized data, leaving open the question of how to order examples at large scale -- i.e. distributed learning with decentralized data. To alleviate the limitation, in this paper we propose D-GraB, an algorithm that orders the examples in a parallel setting with negligible overhead, which enjoys linear speed up at rate $\tilde{O}((mnT)^{-2/3})$ on smooth non-convex objectives and $\tilde{O}((mnT)^{-2})$ under PL condition, where $n$ denotes the number of parallel workers, $m$ denotes the number of examples per worker and $T$ denotes the number of epochs. D-GraB benefits from both data ordering and parallelism. Empirically, we show on various applications including GLUE, CIFAR10 and WikiText-2 that D-GraB outperforms naive parallel GraB and Distributed Random Reshuffling in terms of both training and validation performance.
♻ ☆ Training Subset Selection for Weak Supervision NeurIPS 2022
Existing weak supervision approaches use all the data covered by weak signals to train a classifier. We show both theoretically and empirically that this is not always optimal. Intuitively, there is a tradeoff between the amount of weakly-labeled data and the precision of the weak labels. We explore this tradeoff by combining pretrained data representations with the cut statistic (Muhlenbach et al., 2004) to select (hopefully) high-quality subsets of the weakly-labeled training data. Subset selection applies to any label model and classifier and is very simple to plug in to existing weak supervision pipelines, requiring just a few lines of code. We show our subset selection method improves the performance of weak supervision for a wide range of label models, classifiers, and datasets. Using less weakly-labeled data improves the accuracy of weak supervision pipelines by up to 19% (absolute) on benchmark tasks.
comment: NeurIPS 2022
Multimedia
☆ Neighborhood Contrastive Transformer for Change Captioning
Change captioning is to describe the semantic change between a pair of similar images in natural language. It is more challenging than general image captioning, because it requires capturing fine-grained change information while being immune to irrelevant viewpoint changes, and solving syntax ambiguity in change descriptions. In this paper, we propose a neighborhood contrastive transformer to improve the model's perceiving ability for various changes under different scenes and cognition ability for complex syntax structure. Concretely, we first design a neighboring feature aggregating to integrate neighboring context into each feature, which helps quickly locate the inconspicuous changes under the guidance of conspicuous referents. Then, we devise a common feature distilling to compare two images at neighborhood level and extract common properties from each image, so as to learn effective contrastive information between them. Finally, we introduce the explicit dependencies between words to calibrate the transformer decoder, which helps better understand complex syntax structure during training. Extensive experimental results demonstrate that the proposed method achieves the state-of-the-art performance on three public datasets with different change scenarios. The code is available at https://github.com/tuyunbin/NCT.
comment: Accepted by IEEE TMM
☆ IPA-CLIP: Integrating Phonetic Priors into Vision and Language Pretraining
Recently, large-scale Vision and Language (V\&L) pretraining has become the standard backbone of many multimedia systems. While it has shown remarkable performance even in unseen situations, it often performs in ways not intuitive to humans. Particularly, they usually do not consider the pronunciation of the input, which humans would utilize to understand language, especially when it comes to unknown words. Thus, this paper inserts phonetic prior into Contrastive Language-Image Pretraining (CLIP), one of the V\&L pretrained models, to make it consider the pronunciation similarity among its pronunciation inputs. To achieve this, we first propose a phoneme embedding that utilizes the phoneme relationships provided by the International Phonetic Alphabet (IPA) chart as a phonetic prior. Next, by distilling the frozen CLIP text encoder, we train a pronunciation encoder employing the IPA-based embedding. The proposed model named IPA-CLIP comprises this pronunciation encoder and the original CLIP encoders (image and text). Quantitative evaluation reveals that the phoneme distribution on the embedding space represents phonetic relationships more accurately when using the proposed phoneme embedding. Furthermore, in some multimodal retrieval tasks, we confirm that the proposed pronunciation encoder enhances the performance of the text encoder and that the pronunciation encoder handles nonsense words in a more phonetic manner than the text encoder. Finally, qualitative evaluation verifies the correlation between the pronunciation encoder and human perception regarding pronunciation similarity.
comment: 11 pages, 8 figures, 5 Tables
☆ Video Question Answering Using CLIP-Guided Visual-Text Attention
Cross-modal learning of video and text plays a key role in Video Question Answering (VideoQA). In this paper, we propose a visual-text attention mechanism to utilize the Contrastive Language-Image Pre-training (CLIP) trained on lots of general domain language-image pairs to guide the cross-modal learning for VideoQA. Specifically, we first extract video features using a TimeSformer and text features using a BERT from the target application domain, and utilize CLIP to extract a pair of visual-text features from the general-knowledge domain through the domain-specific learning. We then propose a Cross-domain Learning to extract the attention information between visual and linguistic features across the target domain and general domain. The set of CLIP-guided visual-text features are integrated to predict the answer. The proposed method is evaluated on MSVD-QA and MSRVTT-QA datasets, and outperforms state-of-the-art methods.
☆ Confidence-based Event-centric Online Video Question Answering on a Newly Constructed ATBS Dataset
Deep neural networks facilitate video question answering (VideoQA), but the real-world applications on video streams such as CCTV and live cast place higher demands on the solver. To address the challenges of VideoQA on long videos of unknown length, we define a new set of problems called Online Open-ended Video Question Answering (O^2VQA). It requires an online state-updating mechanism for the solver to decide if the collected information is sufficient to conclude an answer. We then propose a Confidence-based Event-centric Online Video Question Answering (CEO-VQA) model to solve this problem. Furthermore, a dataset called Answer Target in Background Stream (ATBS) is constructed to evaluate this newly developed online VideoQA application. Compared to the baseline VideoQA method that watches the whole video, the experimental results show that the proposed method achieves a significant performance gain.
☆ Butterfly: Multiple Reference Frames Feature Propagation Mechanism for Neural Video Compression
Using more reference frames can significantly improve the compression efficiency in neural video compression. However, in low-latency scenarios, most existing neural video compression frameworks usually use the previous one frame as reference. Or a few frameworks which use the previous multiple frames as reference only adopt a simple multi-reference frames propagation mechanism. In this paper, we present a more reasonable multi-reference frames propagation mechanism for neural video compression, called butterfly multi-reference frame propagation mechanism (Butterfly), which allows a more effective feature fusion of multi-reference frames. By this, we can generate more accurate temporal context conditional prior for Contextual Coding Module. Besides, when the number of decoded frames does not meet the required number of reference frames, we duplicate the nearest reference frame to achieve the requirement, which is better than duplicating the furthest one. Experiment results show that our method can significantly outperform the previous state-of-the-art (SOTA), and our neural codec can achieve -7.6% bitrate save on HEVC Class D dataset when compares with our base single-reference frame model with the same compression configuration.
comment: Accepted by DCC 2023
♻ ☆ Compose & Embellish: Well-Structured Piano Performance Generation via A Two-Stage Approach ICASSP
Even with strong sequence models like Transformers, generating expressive piano performances with long-range musical structures remains challenging. Meanwhile, methods to compose well-structured melodies or lead sheets (melody + chords), i.e., simpler forms of music, gained more success. Observing the above, we devise a two-stage Transformer-based framework that Composes a lead sheet first, and then Embellishes it with accompaniment and expressive touches. Such a factorization also enables pretraining on non-piano data. Our objective and subjective experiments show that Compose & Embellish shrinks the gap in structureness between a current state of the art and real performances by half, and improves other musical aspects such as richness and coherence as well.
comment: Accepted to International Conference on Acoustics, Speech, and Signal Processing (ICASSP) 2023
♻ ☆ Perfectly Secure Steganography Using Minimum Entropy Coupling
Steganography is the practice of encoding secret information into innocuous content in such a manner that an adversarial third party would not realize that there is hidden meaning. While this problem has classically been studied in security literature, recent advances in generative models have led to a shared interest among security and machine learning researchers in developing scalable steganography techniques. In this work, we show that a steganography procedure is perfectly secure under \citet{cachin_perfect}'s information theoretic-model of steganography if and only if it is induced by a coupling. Furthermore, we show that, among perfectly secure procedures, a procedure is maximally efficient if and only if it is induced by a minimum entropy coupling. These insights yield what are, to the best of our knowledge, the first steganography algorithms to achieve perfect security guarantees with non-trivial efficiency; additionally, these algorithms are highly scalable. To provide empirical validation, we compare a minimum entropy coupling-based approach to three modern baselines -- arithmetic coding, Meteor, and adaptive dynamic grouping -- using GPT-2 and WaveRNN as communication channels. We find that the minimum entropy coupling-based approach yields superior encoding efficiency, despite its stronger security constraints. In aggregate, these results suggest that it may be natural to view information-theoretic steganography through the lens of minimum entropy coupling.
♻ ☆ MultiViz: Towards Visualizing and Understanding Multimodal Models ICLR 2023
The promise of multimodal models for real-world applications has inspired research in visualizing and understanding their internal mechanics with the end goal of empowering stakeholders to visualize model behavior, perform model debugging, and promote trust in machine learning models. However, modern multimodal models are typically black-box neural networks, which makes it challenging to understand their internal mechanics. How can we visualize the internal modeling of multimodal interactions in these models? Our paper aims to fill this gap by proposing MultiViz, a method for analyzing the behavior of multimodal models by scaffolding the problem of interpretability into 4 stages: (1) unimodal importance: how each modality contributes towards downstream modeling and prediction, (2) cross-modal interactions: how different modalities relate with each other, (3) multimodal representations: how unimodal and cross-modal interactions are represented in decision-level features, and (4) multimodal prediction: how decision-level features are composed to make a prediction. MultiViz is designed to operate on diverse modalities, models, tasks, and research areas. Through experiments on 8 trained models across 6 real-world tasks, we show that the complementary stages in MultiViz together enable users to (1) simulate model predictions, (2) assign interpretable concepts to features, (3) perform error analysis on model misclassifications, and (4) use insights from error analysis to debug models. MultiViz is publicly available, will be regularly updated with new interpretation tools and metrics, and welcomes inputs from the community.
comment: ICLR 2023. Code available at: https://github.com/pliang279/MultiViz
Computation and Language
☆ WADER at SemEval-2023 Task 9: A Weak-labelling framework for Data augmentation in tExt Regression Tasks
Intimacy is an essential element of human relationships and language is a crucial means of conveying it. Textual intimacy analysis can reveal social norms in different contexts and serve as a benchmark for testing computational models' ability to understand social information. In this paper, we propose a novel weak-labeling strategy for data augmentation in text regression tasks called WADER. WADER uses data augmentation to address the problems of data imbalance and data scarcity and provides a method for data augmentation in cross-lingual, zero-shot tasks. We benchmark the performance of State-of-the-Art pre-trained multilingual language models using WADER and analyze the use of sampling techniques to mitigate bias in data and optimally select augmentation candidates. Our results show that WADER outperforms the baseline model and provides a direction for mitigating data imbalance and scarcity in text regression tasks.
☆ FQP 2.0: Industry Trend Analysis via Hierarchical Financial Data
Analyzing trends across industries is critical to maintaining a healthy and stable economy. Previous research has mainly analyzed official statistics, which are more accurate but not necessarily real-time. In this paper, we propose a method for analyzing industry trends using stock market data. The difficulty of this task is that the raw data is relatively noisy, which affects the accuracy of statistical analysis. In addition, textual data for industry analysis needs to be better understood through language models. For this reason, we introduce the method of industry trend analysis from two perspectives of explicit analysis and implicit analysis. For the explicit analysis, we introduce a hierarchical data (industry and listed company) analysis method to reduce the impact of noise. For implicit analysis, we further pre-train GPT-2 to analyze industry trends with current affairs background as input, making full use of the knowledge learned in the pre-training corpus. We conduct experiments based on the proposed method and achieve good industry trend analysis results.
☆ Mining both Commonality and Specificity from Multiple Documents for Multi-Document Summarization
The multi-document summarization task requires the designed summarizer to generate a short text that covers the important information of original documents and satisfies content diversity. This paper proposes a multi-document summarization approach based on hierarchical clustering of documents. It utilizes the constructed class tree of documents to extract both the sentences reflecting the commonality of all documents and the sentences reflecting the specificity of some subclasses of these documents for generating a summary, so as to satisfy the coverage and diversity requirements of multi-document summarization. Comparative experiments with different variant approaches on DUC'2002-2004 datasets prove the effectiveness of mining both the commonality and specificity of documents for multi-document summarization. Experiments on DUC'2004 and Multi-News datasets show that our approach achieves competitive performance compared to the state-of-the-art unsupervised and supervised approaches.
comment: 10 pages, 1 figure
☆ Knowledge-Based Counterfactual Queries for Visual Question Answering
Visual Question Answering (VQA) has been a popular task that combines vision and language, with numerous relevant implementations in literature. Even though there are some attempts that approach explainability and robustness issues in VQA models, very few of them employ counterfactuals as a means of probing such challenges in a model-agnostic way. In this work, we propose a systematic method for explaining the behavior and investigating the robustness of VQA models through counterfactual perturbations. For this reason, we exploit structured knowledge bases to perform deterministic, optimal and controllable word-level replacements targeting the linguistic modality, and we then evaluate the model's response against such counterfactual inputs. Finally, we qualitatively extract local and global explanations based on counterfactual responses, which are ultimately proven insightful towards interpreting VQA model behaviors. By performing a variety of perturbation types, targeting different parts of speech of the input question, we gain insights to the reasoning of the model, through the comparison of its responses in different adversarial circumstances. Overall, we reveal possible biases in the decision-making process of the model, as well as expected and unexpected patterns, which impact its performance quantitatively and qualitatively, as indicated by our analysis.
☆ Effectiveness of Data Augmentation for Prefix Tuning with Limited Data
Recent work has demonstrated that tuning continuous prompts on large, frozen pretrained language models (i.e., prefix tuning or P-tuning) can yield performance that is comparable or superior to fine-tuning. Nevertheless, the effectiveness of such methods under the context of data augmentation, which has been considered a common strategy to improve learning under low data regimes, has not be studied. In this paper, we examine several popular task-agnostic data augmentation techniques, i.e., EDA, Back Translation, and Mixup, when using prefix tuning under data scarcity. We show that data augmentation can be used to boost the performance of prefix tuning models, but the effectiveness of each technique varies and certain methods can lead to a notable degradation in performance, particularly when using larger models and on harder tasks. To help understand the above behaviour, we run experiments which reveal how prefix tuning generally presents a limited ability to separate the sentence embeddings from different classes of augmented data, and displays poorer performance on heavily altered data in particular. We also demonstrate that by adding a simple contrastive loss we can help mitigate such issues for prefix tuning, resulting in an improvement to augmented data performance.
☆ FinXABSA:Explainable Finance through Aspect-Based Sentiment Analysis
This paper presents a novel approach for explainability in financial analysis by utilizing the Pearson correlation coefficient to establish a relationship between aspect-based sentiment analysis and stock prices. The proposed methodology involves constructing an aspect list from financial news articles and analyzing sentiment intensity scores for each aspect. These scores are then compared to the stock prices for the relevant companies using the Pearson coefficient to determine any significant correlations. The results indicate that the proposed approach provides a more detailed and accurate understanding of the relationship between sentiment analysis and stock prices, which can be useful for investors and financial analysts in making informed decisions. Additionally, this methodology offers a transparent and interpretable way to explain the sentiment analysis results and their impact on stock prices. Overall, the findings of this paper demonstrate the importance of explainability in financial analysis and highlight the potential benefits of utilizing the Pearson coefficient for analyzing aspect-based sentiment analysis and stock prices. The proposed approach offers a valuable tool for understanding the complex relationships between financial news sentiment and stock prices, providing a new perspective on the financial market and aiding in making informed investment decisions.
♻ ☆ A Sidecar Separator Can Convert a Single-Talker Speech Recognition System to a Multi-Talker One ICASSP
Although automatic speech recognition (ASR) can perform well in common non-overlapping environments, sustaining performance in multi-talker overlapping speech recognition remains challenging. Recent research revealed that ASR model's encoder captures different levels of information with different layers -- the lower layers tend to have more acoustic information, and the upper layers more linguistic. This inspires us to develop a Sidecar separator to empower a well-trained ASR model for multi-talker scenarios by separating the mixed speech embedding between two suitable layers. We experimented with a wav2vec 2.0-based ASR model with a Sidecar mounted. By freezing the parameters of the original model and training only the Sidecar (8.7 M, 8.4% of all parameters), the proposed approach outperforms the previous state-of-the-art by a large margin for the 2-speaker mixed LibriMix dataset, reaching a word error rate (WER) of 10.36%; and obtains comparable results (7.56%) for LibriSpeechMix dataset when limited training.
comment: Accepted by IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), 2023
♻ ☆ E2E Segmentation in a Two-Pass Cascaded Encoder ASR Model ICASSP 2023
We explore unifying a neural segmenter with two-pass cascaded encoder ASR into a single model. A key challenge is allowing the segmenter (which runs in real-time, synchronously with the decoder) to finalize the 2nd pass (which runs 900 ms behind real-time) without introducing user-perceived latency or deletion errors during inference. We propose a design where the neural segmenter is integrated with the causal 1st pass decoder to emit a end-of-segment (EOS) signal in real-time. The EOS signal is then used to finalize the non-causal 2nd pass. We experiment with different ways to finalize the 2nd pass, and find that a novel dummy frame injection strategy allows for simultaneous high quality 2nd pass results and low finalization latency. On a real-world long-form captioning task (YouTube), we achieve 2.4% relative WER and 140 ms EOS latency gains over a baseline VAD-based segmenter with the same cascaded encoder.
comment: ICASSP 2023
♻ ☆ Advancements in Federated Learning: Models, Methods, and Privacy
Federated learning (FL) is a promising technique for addressing the rising privacy and security issues. Its main ingredient is to cooperatively learn the model among the distributed clients without uploading any sensitive data. In this paper, we conducted a thorough review of the related works, following the development context and deeply mining the key technologies behind FL from both theoretical and practical perspectives. Specifically, we first classify the existing works in FL architecture based on the network topology of FL systems with detailed analysis and summarization. Next, we abstract the current application problems, summarize the general techniques and frame the application problems into the general paradigm of FL base models. Moreover, we provide our proposed solutions for model training via FL. We have summarized and analyzed the existing FedOpt algorithms, and deeply revealed the algorithmic development principles of many first-order algorithms in depth, proposing a more generalized algorithm design framework. Based on these frameworks, we have instantiated FedOpt algorithms. As privacy and security is the fundamental requirement in FL, we provide the existing attack scenarios and the defense methods. To the best of our knowledge, we are among the first tier to review the theoretical methodology and propose our strategies since there are very few works surveying the theoretical approaches. Our survey targets motivating the development of high-performance, privacy-preserving, and secure methods to integrate FL into real-world applications.
comment: 35 pages, submitted to ACM Computing Surveys
♻ ☆ Looking for a Needle in a Haystack: A Comprehensive Study of Hallucinations in Neural Machine Translation EACL23
Although the problem of hallucinations in neural machine translation (NMT) has received some attention, research on this highly pathological phenomenon lacks solid ground. Previous work has been limited in several ways: it often resorts to artificial settings where the problem is amplified, it disregards some (common) types of hallucinations, and it does not validate adequacy of detection heuristics. In this paper, we set foundations for the study of NMT hallucinations. First, we work in a natural setting, i.e., in-domain data without artificial noise neither in training nor in inference. Next, we annotate a dataset of over 3.4k sentences indicating different kinds of critical errors and hallucinations. Then, we turn to detection methods and both revisit methods used previously and propose using glass-box uncertainty-based detectors. Overall, we show that for preventive settings, (i) previously used methods are largely inadequate, (ii) sequence log-probability works best and performs on par with reference-based methods. Finally, we propose DeHallucinator, a simple method for alleviating hallucinations at test time that significantly reduces the hallucinatory rate. To ease future research, we release our annotated dataset for WMT18 German-English data, along with the model, training data, and code.
comment: Accepted at EACL23 (main)
♻ ☆ AudioGen: Textually Guided Audio Generation ICLR 2023
We tackle the problem of generating audio samples conditioned on descriptive text captions. In this work, we propose AaudioGen, an auto-regressive generative model that generates audio samples conditioned on text inputs. AudioGen operates on a learnt discrete audio representation. The task of text-to-audio generation poses multiple challenges. Due to the way audio travels through a medium, differentiating ``objects'' can be a difficult task (e.g., separating multiple people simultaneously speaking). This is further complicated by real-world recording conditions (e.g., background noise, reverberation, etc.). Scarce text annotations impose another constraint, limiting the ability to scale models. Finally, modeling high-fidelity audio requires encoding audio at high sampling rate, leading to extremely long sequences. To alleviate the aforementioned challenges we propose an augmentation technique that mixes different audio samples, driving the model to internally learn to separate multiple sources. We curated 10 datasets containing different types of audio and text annotations to handle the scarcity of text-audio data points. For faster inference, we explore the use of multi-stream modeling, allowing the use of shorter sequences while maintaining a similar bitrate and perceptual quality. We apply classifier-free guidance to improve adherence to text. Comparing to the evaluated baselines, AudioGen outperforms over both objective and subjective metrics. Finally, we explore the ability of the proposed method to generate audio continuation conditionally and unconditionally. Samples: https://felixkreuk.github.io/audiogen
comment: Accepted to ICLR 2023
♻ ☆ Chain of Hindsight Aligns Language Models with Feedback
Learning from human preferences is important for language models to be helpful and useful for humans, and to align with human and social values. Prior work have achieved remarkable successes by learning from human feedback to understand and follow instructions. Nonetheless, these methods are either founded on hand-picked model generations that are favored by human annotators, rendering them ineffective in terms of data utilization and challenging to apply in general, or they depend on reward functions and reinforcement learning, which are prone to imperfect reward function and extremely challenging to optimize. In this work, we propose a novel technique, Chain of Hindsight, that is easy to optimize and can learn from any form of feedback, regardless of its polarity. Our idea is inspired by how humans learn from extensive feedback presented in the form of languages. We convert all types of feedback into sentences, which are then used to fine-tune the model, allowing us to take advantage of the language comprehension capabilities of language models. We condition the model on a sequence of model generations paired with feedback. By doing so, models are trained to generate outputs based on feedback, and models can learn to identify and correct negative attributes or errors. Applying our method to large language models, we observed that Chain of Hindsight significantly surpasses previous methods in aligning language models with human preferences. We observed significant improvements on summarization and dialogue tasks and our approach is markedly preferred in human evaluations.
comment: Included a comprehensive list of templates
Computer Vision and Pattern Recognition
☆ DeepMAD: Mathematical Architecture Design for Deep Convolutional Neural Network CVPR 2023
The rapid advances in Vision Transformer (ViT) refresh the state-of-the-art performances in various vision tasks, overshadowing the conventional CNN-based models. This ignites a few recent striking-back research in the CNN world showing that pure CNN models can achieve as good performance as ViT models when carefully tuned. While encouraging, designing such high-performance CNN models is challenging, requiring non-trivial prior knowledge of network design. To this end, a novel framework termed Mathematical Architecture Design for Deep CNN (DeepMAD) is proposed to design high-performance CNN models in a principled way. In DeepMAD, a CNN network is modeled as an information processing system whose expressiveness and effectiveness can be analytically formulated by their structural parameters. Then a constrained mathematical programming (MP) problem is proposed to optimize these structural parameters. The MP problem can be easily solved by off-the-shelf MP solvers on CPUs with a small memory footprint. In addition, DeepMAD is a pure mathematical framework: no GPU or training data is required during network design. The superiority of DeepMAD is validated on multiple large-scale computer vision benchmark datasets. Notably on ImageNet-1k, only using conventional convolutional layers, DeepMAD achieves 0.7% and 1.5% higher top-1 accuracy than ConvNeXt and Swin on Tiny level, and 0.8% and 0.9% higher on Small level.
comment: Accepted by CVPR 2023
☆ A Low-Cost Portable Apparatus to Analyze Oral Fluid Droplets and Quantify the Efficacy of Masks
Every year, about 4 million people die from upper respiratory infections. Mask-wearing is crucial in preventing the spread of pathogen-containing droplets, which is the primary cause of these illnesses. However, most techniques for mask efficacy evaluation are expensive to set up and complex to operate. In this work, a novel, low-cost, and quantitative metrology to visualize, track, and analyze orally-generated fluid droplets is developed. The project has four stages: setup optimization, data collection, data analysis, and application development. The metrology was initially developed in a dark closet as a proof of concept using common household materials and was subsequently implemented into a portable apparatus. Tonic water and UV darklight tube lights are selected to visualize fluorescent droplet and aerosol propagation with automated analysis developed using open-source software. The dependencies of oral fluid droplet generation and propagation on various factors are studied in detail and established using this metrology. Additionally, the smallest detectable droplet size was mathematically correlated to height and airborne time. The efficacy of different types of masks is evaluated and associated with fabric microstructures. It is found that masks with smaller-sized pores and thicker material are more effective. This technique can easily be constructed at home using materials that total to a cost of below \$60, thereby enabling a low-cost and accurate metrology.
comment: 13 pages, 15 figures. arXiv admin note: substantial text overlap with arXiv:2201.03993
☆ A Study of Augmentation Methods for Handwritten Stenography Recognition
One of the factors limiting the performance of handwritten text recognition (HTR) for stenography is the small amount of annotated training data. To alleviate the problem of data scarcity, modern HTR methods often employ data augmentation. However, due to specifics of the stenographic script, such settings may not be directly applicable for stenography recognition. In this work, we study 22 classical augmentation techniques, most of which are commonly used for HTR of other scripts, such as Latin handwriting. Through extensive experiments, we identify a group of augmentations, including for example contained ranges of random rotation, shifts and scaling, that are beneficial to the use case of stenography recognition. Furthermore, a number of augmentation approaches, leading to a decrease in recognition performance, are identified. Our results are supported by statistical hypothesis testing. Links to the publicly available dataset and codebase are provided.
☆ Human-Art: A Versatile Human-Centric Dataset Bridging Natural and Artificial Scenes CVPR2023
Humans have long been recorded in a variety of forms since antiquity. For example, sculptures and paintings were the primary media for depicting human beings before the invention of cameras. However, most current human-centric computer vision tasks like human pose estimation and human image generation focus exclusively on natural images in the real world. Artificial humans, such as those in sculptures, paintings, and cartoons, are commonly neglected, making existing models fail in these scenarios. As an abstraction of life, art incorporates humans in both natural and artificial scenes. We take advantage of it and introduce the Human-Art dataset to bridge related tasks in natural and artificial scenarios. Specifically, Human-Art contains 50k high-quality images with over 123k person instances from 5 natural and 15 artificial scenarios, which are annotated with bounding boxes, keypoints, self-contact points, and text information for humans represented in both 2D and 3D. It is, therefore, comprehensive and versatile for various downstream tasks. We also provide a rich set of baseline results and detailed analyses for related tasks, including human detection, 2D and 3D human pose estimation, image generation, and motion transfer. As a challenging dataset, we hope Human-Art can provide insights for relevant research and open up new research questions.
comment: CVPR2023
☆ Frequency-domain Blind Quality Assessment of Blurred and Blocking-artefact Images using Gaussian Process Regression model
Most of the standard image and video codecs are block-based and depending upon the compression ratio the compressed images/videos suffer from different distortions. At low ratios, blurriness is observed and as compression increases blocking artifacts occur. Generally, in order to reduce blockiness, images are low-pass filtered which leads to more blurriness. Also, in bokeh mode images they are commonly seen: blurriness as a result of intentional blurred background while blocking artifact and global blurriness arising due to compression. Therefore, such visual media suffer from both blockiness and blurriness distortions. Along with this, noise is also commonly encountered distortion. Most of the existing works on quality assessment quantify these distortions individually. This paper proposes a methodology to blindly measure overall quality of an image suffering from these distortions, individually as well as jointly. This is achieved by considering the sum of absolute values of low and high-frequency Discrete Frequency Transform (DFT) coefficients defined as sum magnitudes. The number of blocks lying in specific ranges of sum magnitudes including zero-valued AC coefficients and mean of 100 maximum and 100 minimum values of these sum magnitudes are used as feature vectors. These features are then fed to the Machine Learning (ML) based Gaussian Process Regression (GPR) model, which quantifies the image quality. The simulation results show that the proposed method can estimate the quality of images distorted with the blockiness, blurriness, noise and their combinations. It is relatively fast compared to many state-of-art methods, and therefore is suitable for real-time quality monitoring applications.
☆ IDA: Informed Domain Adaptive Semantic Segmentation
Mixup-based data augmentation has been validated to be a critical stage in the self-training framework for unsupervised domain adaptive semantic segmentation (UDA-SS), which aims to transfer knowledge from a well-annotated (source) domain to an unlabeled (target) domain. Existing self-training methods usually adopt the popular region-based mixup techniques with a random sampling strategy, which unfortunately ignores the dynamic evolution of different semantics across various domains as training proceeds. To improve the UDA-SS performance, we propose an Informed Domain Adaptation (IDA) model, a self-training framework that mixes the data based on class-level segmentation performance, which aims to emphasize small-region semantics during mixup. In our IDA model, the class-level performance is tracked by an expected confidence score (ECS). We then use a dynamic schedule to determine the mixing ratio for data in different domains. Extensive experimental results reveal that our proposed method is able to outperform the state-of-the-art UDA-SS method by a margin of 1.1 mIoU in the adaptation of GTA-V to Cityscapes and of 0.9 mIoU in the adaptation of SYNTHIA to Cityscapes.
☆ SePaint: Semantic Map Inpainting via Multinomial Diffusion
Prediction beyond partial observations is crucial for robots to navigate in unknown environments because it can provide extra information regarding the surroundings beyond the current sensing range or resolution. In this work, we consider the inpainting of semantic Bird's-Eye-View maps. We propose SePaint, an inpainting model for semantic data based on generative multinomial diffusion. To maintain semantic consistency, we need to condition the prediction for the missing regions on the known regions. We propose a novel and efficient condition strategy, Look-Back Condition (LB-Con), which performs one-step look-back operations during the reverse diffusion process. By doing so, we are able to strengthen the harmonization between unknown and known parts, leading to better completion performance. We have conducted extensive experiments on different datasets, showing our proposed model outperforms commonly used interpolation methods in various robotic applications.
☆ Scalable Object Detection on Embedded Devices Using Weight Pruning and Singular Value Decomposition
This paper presents a method for optimizing object detection models by combining weight pruning and singular value decomposition (SVD). The proposed method was evaluated on a custom dataset of street work images obtained from https://universe.roboflow.com/roboflow-100/street-work. The dataset consists of 611 training images, 175 validation images, and 87 test images with 7 classes. We compared the performance of the optimized models with the original unoptimized model in terms of frame rate, mean average precision (mAP@50), and weight size. The results show that the weight pruning + SVD model achieved a 0.724 mAP@50 with a frame rate of 1.48 FPS and a weight size of 12.1 MB, outperforming the original model (0.717 mAP@50, 1.50 FPS, and 12.3 MB). Precision-recall curves were also plotted for all models. Our work demonstrates that the proposed method can effectively optimize object detection models while balancing accuracy, speed, and model size.
comment: 8 pages, 3 figures. A report of the project done as part of the Yonsei-Roboin project for the 2nd semester, 2022
☆ Reparameterization through Spatial Gradient Scaling ICLR 2023
Reparameterization aims to improve the generalization of deep neural networks by transforming convolutional layers into equivalent multi-branched structures during training. However, there exists a gap in understanding how reparameterization may change and benefit the learning process of neural networks. In this paper, we present a novel spatial gradient scaling method to redistribute learning focus among weights in convolutional networks. We prove that spatial gradient scaling achieves the same learning dynamics as a branched reparameterization yet without introducing structural changes into the network. We further propose an analytical approach that dynamically learns scalings for each convolutional layer based on the spatial characteristics of its input feature map gauged by mutual information. Experiments on CIFAR-10, CIFAR-100, and ImageNet show that without searching for reparameterized structures, our proposed scaling method outperforms the state-of-the-art reparameterization strategies at a lower computational cost.
comment: Published at ICLR 2023. Code available $\href{https://github.com/Ascend-Research/Reparameterization}{here}$
☆ Vision based Virtual Guidance for Navigation
This paper explores the impact of virtual guidance on mid-level representation-based navigation, where an agent performs navigation tasks based solely on visual observations. Instead of providing distance measures or numerical directions to guide the agent, which may be difficult for it to interpret visually, the paper investigates the potential of different forms of virtual guidance schemes on navigation performance. Three schemes of virtual guidance signals are explored: virtual navigation path, virtual waypoints, and a combination of both. The experiments were conducted using a virtual city built with the Unity engine to train the agents while avoiding obstacles. The results show that virtual guidance provides the agent with more meaningful navigation information and achieves better performance in terms of path completion rates and navigation efficiency. In addition, a set of analyses were provided to investigate the failure cases and the navigated trajectories, and a pilot study was conducted for the real-world scenarios.
comment: Yu-Ying Chen, Tsung-Chih Chiang, Chia-Chuan Hsu, Chun-Chia Huang, Chun-Wei Huang, Jou-Min Liu, and Ting-Ru Liu contributed equally to this work, names listed in alphabetical order; This work has been submitted to the IEEE for possible publication
☆ Learning to Localize in Unseen Scenes with Relative Pose Regressors
Relative pose regressors (RPRs) localize a camera by estimating its relative translation and rotation to a pose-labelled reference. Unlike scene coordinate regression and absolute pose regression methods, which learn absolute scene parameters, RPRs can (theoretically) localize in unseen environments, since they only learn the residual pose between camera pairs. In practice, however, the performance of RPRs is significantly degraded in unseen scenes. In this work, we propose to aggregate paired feature maps into latent codes, instead of operating on global image descriptors, in order to improve the generalization of RPRs. We implement aggregation with concatenation, projection, and attention operations (Transformer Encoders) and learn to regress the relative pose parameters from the resulting latent codes. We further make use of a recently proposed continuous representation of rotation matrices, which alleviates the limitations of the commonly used quaternions. Compared to state-of-the-art RPRs, our model is shown to localize significantly better in unseen environments, across both indoor and outdoor benchmarks, while maintaining competitive performance in seen scenes. We validate our findings and architecture design through multiple ablations. Our code and pretrained models is publicly available.
☆ Deep Learning in the Field of Biometric Template Protection: An Overview
Today, deep learning represents the most popular and successful form of machine learning. Deep learning has revolutionised the field of pattern recognition, including biometric recognition. Biometric systems utilising deep learning have been shown to achieve auspicious recognition accuracy, surpassing human performance. Apart from said breakthrough advances in terms of biometric performance, the use of deep learning was reported to impact different covariates of biometrics such as algorithmic fairness, vulnerability to attacks, or template protection. Technologies of biometric template protection are designed to enable a secure and privacy-preserving deployment of biometrics. In the recent past, deep learning techniques have been frequently applied in biometric template protection systems for various purposes. This work provides an overview of how advances in deep learning take influence on the field of biometric template protection. The interrelation between improved biometric performance rates and security in biometric template protection is elaborated. Further, the use of deep learning for obtaining feature representations that are suitable for biometric template protection is discussed. Novel methods that apply deep learning to achieve various goals of biometric template protection are surveyed along with deep learning-based attacks.
☆ HairStep: Transfer Synthetic to Real Using Strand and Depth Maps for Single-View 3D Hair Modeling CVPR 2023
In this work, we tackle the challenging problem of learning-based single-view 3D hair modeling. Due to the great difficulty of collecting paired real image and 3D hair data, using synthetic data to provide prior knowledge for real domain becomes a leading solution. This unfortunately introduces the challenge of domain gap. Due to the inherent difficulty of realistic hair rendering, existing methods typically use orientation maps instead of hair images as input to bridge the gap. We firmly think an intermediate representation is essential, but we argue that orientation map using the dominant filtering-based methods is sensitive to uncertain noise and far from a competent representation. Thus, we first raise this issue up and propose a novel intermediate representation, termed as HairStep, which consists of a strand map and a depth map. It is found that HairStep not only provides sufficient information for accurate 3D hair modeling, but also is feasible to be inferred from real images. Specifically, we collect a dataset of 1,250 portrait images with two types of annotations. A learning framework is further designed to transfer real images to the strand map and depth map. It is noted that, an extra bonus of our new dataset is the first quantitative metric for 3D hair modeling. Our experiments show that HairStep narrows the domain gap between synthetic and real and achieves state-of-the-art performance on single-view 3D hair reconstruction.
comment: CVPR 2023, project page: https://paulyzheng.github.io/research/hairstep/
☆ Robust affine feature matching via quadratic assignment on Grassmannians
GraNNI (Grassmannians for Nearest Neighbours Identification) a new algorithm to solve the problem of affine registration is proposed. The algorithm is based on the Grassmannian of $k$--dimensional planes in $\mathbb{R}^n$ and minimizing the Frobenius norm between the two elements of the Grassmannian. The Quadratic Assignment Problem (QAP) is used to find the matching. The results of the experiments show that the algorithm is more robust to noise and point discrepancy in point clouds than previous approaches.
☆ Maximizing Spatio-Temporal Entropy of Deep 3D CNNs for Efficient Video Recognition ICLR 2023
3D convolution neural networks (CNNs) have been the prevailing option for video recognition. To capture the temporal information, 3D convolutions are computed along the sequences, leading to cubically growing and expensive computations. To reduce the computational cost, previous methods resort to manually designed 3D/2D CNN structures with approximations or automatic search, which sacrifice the modeling ability or make training time-consuming. In this work, we propose to automatically design efficient 3D CNN architectures via a novel training-free neural architecture search approach tailored for 3D CNNs considering the model complexity. To measure the expressiveness of 3D CNNs efficiently, we formulate a 3D CNN as an information system and derive an analytic entropy score, based on the Maximum Entropy Principle. Specifically, we propose a spatio-temporal entropy score (STEntr-Score) with a refinement factor to handle the discrepancy of visual information in spatial and temporal dimensions, through dynamically leveraging the correlation between the feature map size and kernel size depth-wisely. Highly efficient and expressive 3D CNN architectures, \ie entropy-based 3D CNNs (E3D family), can then be efficiently searched by maximizing the STEntr-Score under a given computational budget, via an evolutionary algorithm without training the network parameters. Extensive experiments on Something-Something V1\&V2 and Kinetics400 demonstrate that the E3D family achieves state-of-the-art performance with higher computational efficiency. Code is available at https://github.com/alibaba/lightweight-neural-architecture-search.
comment: This manuscript has been accepted at ICLR 2023
☆ Text2Face: A Multi-Modal 3D Face Model
We present the first 3D morphable modelling approach, whereby 3D face shape can be directly and completely defined using a textual prompt. Building on work in multi-modal learning, we extend the FLAME head model to a common image-and-text latent space. This allows for direct 3D Morphable Model (3DMM) parameter generation and therefore shape manipulation from textual descriptions. Our method, Text2Face, has many applications; for example: generating police photofits where the input is already in natural language. It further enables multi-modal 3DMM image fitting to sketches and sculptures, as well as images.
☆ Continuous-Time Gaussian Process Motion-Compensation for Event-vision Pattern Tracking with Distance Fields
This work addresses the issue of motion compensation and pattern tracking in event camera data. An event camera generates asynchronous streams of events triggered independently by each of the pixels upon changes in the observed intensity. Providing great advantages in low-light and rapid-motion scenarios, such unconventional data present significant research challenges as traditional vision algorithms are not directly applicable to this sensing modality. The proposed method decomposes the tracking problem into a local SE(2) motion-compensation step followed by a homography registration of small motion-compensated event batches. The first component relies on Gaussian Process (GP) theory to model the continuous occupancy field of the events in the image plane and embed the camera trajectory in the covariance kernel function. In doing so, estimating the trajectory is done similarly to GP hyperparameter learning by maximising the log marginal likelihood of the data. The continuous occupancy fields are turned into distance fields and used as templates for homography-based registration. By benchmarking the proposed method against other state-of-the-art techniques, we show that our open-source implementation performs high-accuracy motion compensation and produces high-quality tracks in real-world scenarios.
comment: Accepted for presentation at the 2023 IEEE International Conference on Robotics and Automation
☆ Learned Lossless Compression for JPEG via Frequency-Domain Prediction
JPEG images can be further compressed to enhance the storage and transmission of large-scale image datasets. Existing learned lossless compressors for RGB images cannot be well transferred to JPEG images due to the distinguishing distribution of DCT coefficients and raw pixels. In this paper, we propose a novel framework for learned lossless compression of JPEG images that achieves end-to-end optimized prediction of the distribution of decoded DCT coefficients. To enable learning in the frequency domain, DCT coefficients are partitioned into groups to utilize implicit local redundancy. An autoencoder-like architecture is designed based on the weight-shared blocks to realize entropy modeling of grouped DCT coefficients and independently compress the priors. We attempt to realize learned lossless compression of JPEG images in the frequency domain. Experimental results demonstrate that the proposed framework achieves superior or comparable performance in comparison to most recent lossless compressors with handcrafted context modeling for JPEG images.
☆ SynthASpoof: Developing Face Presentation Attack Detection Based on Privacy-friendly Synthetic Data
Recently, significant progress has been made in face presentation attack detection (PAD), which aims to secure face recognition systems against presentation attacks, owing to the availability of several face PAD datasets. However, all available datasets are based on privacy and legally-sensitive authentic biometric data with a limited number of subjects. To target these legal and technical challenges, this work presents the first synthetic-based face PAD dataset, named SynthASpoof, as a large-scale PAD development dataset. The bona fide samples in SynthASpoof are synthetically generated and the attack samples are collected by presenting such synthetic data to capture systems in a real attack scenario. The experimental results demonstrate the feasibility of using SynthASpoof for the development of face PAD. Moreover, we boost the performance of such a solution by incorporating the domain generalization tool MixStyle into the PAD solutions. Additionally, we showed the viability of using synthetic data as a supplement to enrich the diversity of limited authentic training data and consistently enhance PAD performances. The SynthASpoof dataset, containing 25,000 bona fide and 78,800 attack samples, the implementation, and the pre-trained weights are made publicly available.
☆ On Modifying a Neural Network's Perception
Artificial neural networks have proven to be extremely useful models that have allowed for multiple recent breakthroughs in the field of Artificial Intelligence and many others. However, they are typically regarded as black boxes, given how difficult it is for humans to interpret how these models reach their results. In this work, we propose a method which allows one to modify what an artificial neural network is perceiving regarding specific human-defined concepts, enabling the generation of hypothetical scenarios that could help understand and even debug the neural network model. Through empirical evaluation, in a synthetic dataset and in the ImageNet dataset, we test the proposed method on different models, assessing whether the performed manipulations are well interpreted by the models, and analyzing how they react to them.
☆ Comparative study of Transformer and LSTM Network with attention mechanism on Image Captioning
In a globalized world at the present epoch of generative intelligence, most of the manual labour tasks are automated with increased efficiency. This can support businesses to save time and money. A crucial component of generative intelligence is the integration of vision and language. Consequently, image captioning become an intriguing area of research. There have been multiple attempts by the researchers to solve this problem with different deep learning architectures, although the accuracy has increased, but the results are still not up to standard. This study buckles down to the comparison of Transformer and LSTM with attention block model on MS-COCO dataset, which is a standard dataset for image captioning. For both the models we have used pretrained Inception-V3 CNN encoder for feature extraction of the images. The Bilingual Evaluation Understudy score (BLEU) is used to checked the accuracy of caption generated by both models. Along with the transformer and LSTM with attention block models,CLIP-diffusion model, M2-Transformer model and the X-Linear Attention model have been discussed with state of the art accuracy.
comment: 13 pages, 7 figures, 2 tables
☆ CueCAn: Cue Driven Contextual Attention For Identifying Missing Traffic Signs on Unconstrained Roads ICRA'23
Unconstrained Asian roads often involve poor infrastructure, affecting overall road safety. Missing traffic signs are a regular part of such roads. Missing or non-existing object detection has been studied for locating missing curbs and estimating reasonable regions for pedestrians on road scene images. Such methods involve analyzing task-specific single object cues. In this paper, we present the first and most challenging video dataset for missing objects, with multiple types of traffic signs for which the cues are visible without the signs in the scenes. We refer to it as the Missing Traffic Signs Video Dataset (MTSVD). MTSVD is challenging compared to the previous works in two aspects i) The traffic signs are generally not present in the vicinity of their cues, ii) The traffic signs cues are diverse and unique. Also, MTSVD is the first publicly available missing object dataset. To train the models for identifying missing signs, we complement our dataset with 10K traffic sign tracks, with 40 percent of the traffic signs having cues visible in the scenes. For identifying missing signs, we propose the Cue-driven Contextual Attention units (CueCAn), which we incorporate in our model encoder. We first train the encoder to classify the presence of traffic sign cues and then train the entire segmentation model end-to-end to localize missing traffic signs. Quantitative and qualitative analysis shows that CueCAn significantly improves the performance of base models.
comment: International Conference on Robotics and Automation (ICRA'23)
☆ VTQA: Visual Text Question Answering via Entity Alignment and Cross-Media Reasoning
The ideal form of Visual Question Answering requires understanding, grounding and reasoning in the joint space of vision and language and serves as a proxy for the AI task of scene understanding. However, most existing VQA benchmarks are limited to just picking the answer from a pre-defined set of options and lack attention to text. We present a new challenge with a dataset that contains 23,781 questions based on 10124 image-text pairs. Specifically, the task requires the model to align multimedia representations of the same entity to implement multi-hop reasoning between image and text and finally use natural language to answer the question. The aim of this challenge is to develop and benchmark models that are capable of multimedia entity alignment, multi-step reasoning and open-ended answer generation.
☆ Deep-Learning-based Counting Methods, Datasets, and Applications in Agriculture -- A Review
The number of objects is considered an important factor in a variety of tasks in the agricultural domain. Automated counting can improve farmers decisions regarding yield estimation, stress detection, disease prevention, and more. In recent years, deep learning has been increasingly applied to many agriculture-related applications, complementing conventional computer-vision algorithms for counting agricultural objects. This article reviews progress in the past decade and the state of the art for counting methods in agriculture, focusing on deep-learning methods. It presents an overview of counting algorithms, metrics, platforms, and sensors, a list of all publicly available datasets, and an in-depth discussion of various deep-learning methods used for counting. Finally, it discusses open challenges in object counting using deep learning and gives a glimpse into new directions and future perspectives for counting research. The review reveals a major leap forward in object counting in agriculture in the past decade, led by the penetration of deep learning methods into counting platforms.
♻ ☆ TubeFormer-DeepLab: Video Mask Transformer CVPR 2022
We present TubeFormer-DeepLab, the first attempt to tackle multiple core video segmentation tasks in a unified manner. Different video segmentation tasks (e.g., video semantic/instance/panoptic segmentation) are usually considered as distinct problems. State-of-the-art models adopted in the separate communities have diverged, and radically different approaches dominate in each task. By contrast, we make a crucial observation that video segmentation tasks could be generally formulated as the problem of assigning different predicted labels to video tubes (where a tube is obtained by linking segmentation masks along the time axis) and the labels may encode different values depending on the target task. The observation motivates us to develop TubeFormer-DeepLab, a simple and effective video mask transformer model that is widely applicable to multiple video segmentation tasks. TubeFormer-DeepLab directly predicts video tubes with task-specific labels (either pure semantic categories, or both semantic categories and instance identities), which not only significantly simplifies video segmentation models, but also advances state-of-the-art results on multiple video segmentation benchmarks
comment: CVPR 2022; arXiv v2: add results on VIPSeg val/test sets and VSPW new test set
♻ ☆ Prediction of Gender from Longitudinal MRI data via Deep Learning on Adolescent Data Reveals Unique Patterns Associated with Brain Structure and Change over a Two-year Period
Deep learning algorithms for predicting neuroimaging data have shown considerable promise in various applications. Prior work has demonstrated that deep learning models that take advantage of the data's 3D structure can outperform standard machine learning on several learning tasks. However, most prior research in this area has focused on neuroimaging data from adults. Within the Adolescent Brain and Cognitive Development (ABCD) dataset, a large longitudinal development study, we examine structural MRI data to predict gender and identify gender-related changes in brain structure. Results demonstrate that gender prediction accuracy is exceptionally high (>97%) with training epochs >200 and that this accuracy increases with age. Brain regions identified as the most discriminative in the task under study include predominantly frontal areas and the temporal lobe. When evaluating gender predictive changes specific to a two-year increase in age, a broader set of visual, cingulate, and insular regions are revealed. Our findings show a robust gender-related structural brain change pattern, even over a small age range. This suggests that it might be possible to study how the brain changes during adolescence by looking at how these changes are related to different behavioral and environmental factors.
comment: I submitted the wrong paper
♻ ☆ MultiCrossViT: Multimodal Vision Transformer for Schizophrenia Prediction using Structural MRI and Functional Network Connectivity Data
Vision Transformer (ViT) is a pioneering deep learning framework that can address real-world computer vision issues, such as image classification and object recognition. Importantly, ViTs are proven to outperform traditional deep learning models, such as convolutional neural networks (CNNs). Relatively recently, a number of ViT mutations have been transplanted into the field of medical imaging, thereby resolving a variety of critical classification and segmentation challenges, especially in terms of brain imaging data. In this work, we provide a novel multimodal deep learning pipeline, MultiCrossViT, which is capable of analyzing both structural MRI (sMRI) and static functional network connectivity (sFNC) data for the prediction of schizophrenia disease. On a dataset with minimal training subjects, our novel model can achieve an AUC of 0.832. Finally, we visualize multiple brain regions and covariance patterns most relevant to schizophrenia based on the resulting ViT attention maps by extracting features from transformer encoders.
comment: I submitted the wrong paper
♻ ☆ MSED: a multi-modal sleep event detection model for clinical sleep analysis
Clinical sleep analysis require manual analysis of sleep patterns for correct diagnosis of sleep disorders. However, several studies have shown significant variability in manual scoring of clinically relevant discrete sleep events, such as arousals, leg movements, and sleep disordered breathing (apneas and hypopneas). We investigated whether an automatic method could be used for event detection and if a model trained on all events (joint model) performed better than corresponding event-specific models (single-event models). We trained a deep neural network event detection model on 1653 individual recordings and tested the optimized model on 1000 separate hold-out recordings. F1 scores for the optimized joint detection model were 0.70, 0.63, and 0.62 for arousals, leg movements, and sleep disordered breathing, respectively, compared to 0.65, 0.61, and 0.60 for the optimized single-event models. Index values computed from detected events correlated positively with manual annotations ($r^2$ = 0.73, $r^2$ = 0.77, $r^2$ = 0.78, respectively). We furthermore quantified model accuracy based on temporal difference metrics, which improved overall by using the joint model compared to single-event models. Our automatic model jointly detects arousals, leg movements and sleep disordered breathing events with high correlation with human annotations. Finally, we benchmark against previous state-of-the-art multi-event detection models and found an overall increase in F1 score with our proposed model despite a 97.5% reduction in model size. Source code for training and inference is available at https://github.com/neergaard/msed.git.
comment: 10 pages, 4 figures. Accepted for publication in IEEE Transactions on Biomedical Engineering
♻ ☆ FewSOL: A Dataset for Few-Shot Object Learning in Robotic Environments
We introduce the Few-Shot Object Learning (FewSOL) dataset for object recognition with a few images per object. We captured 336 real-world objects with 9 RGB-D images per object from different views. Object segmentation masks, object poses and object attributes are provided. In addition, synthetic images generated using 330 3D object models are used to augment the dataset. We investigated (i) few-shot object classification and (ii) joint object segmentation and few-shot classification with the state-of-the-art methods for few-shot learning and meta-learning using our dataset. The evaluation results show that there is still a large margin to be improved for few-shot object classification in robotic environments. Our dataset can be used to study a set of few-shot object recognition problems such as classification, detection and segmentation, shape reconstruction, pose estimation, keypoint correspondences and attribute recognition. The dataset and code are available at https://irvlutd.github.io/FewSOL.
♻ ☆ Polyline Generative Navigable Space Segmentation for Autonomous Visual Navigation
Detecting navigable space is a fundamental capability for mobile robots navigating in unknown or unmapped environments. In this work, we treat visual navigable space segmentation as a scene decomposition problem and propose Polyline Segmentation Variational autoencoder Network (PSV-Net), a representation learning-based framework for learning the navigable space segmentation in a self-supervised manner. Current segmentation techniques heavily rely on fully-supervised learning strategies which demand a large amount of pixel-level annotated images. In this work, we propose a framework leveraging a Variational AutoEncoder (VAE) and an AutoEncoder (AE) to learn a polyline representation that compactly outlines the desired navigable space boundary. Through extensive experiments, we validate that the proposed PSV-Net can learn the visual navigable space with no or few labels, producing an accuracy comparable to fully-supervised state-of-the-art methods that use all available labels. In addition, we show that integrating the proposed navigable space segmentation model with a visual planner can achieve efficient mapless navigation in real environments.
♻ ☆ Decoding natural image stimuli from fMRI data with a surface-based convolutional network
Due to the low signal-to-noise ratio and limited resolution of functional MRI data, and the high complexity of natural images, reconstructing a visual stimulus from human brain fMRI measurements is a challenging task. In this work, we propose a novel approach for this task, which we call Cortex2Image, to decode visual stimuli with high semantic fidelity and rich fine-grained detail. In particular, we train a surface-based convolutional network model that maps from brain response to semantic image features first (Cortex2Semantic). We then combine this model with a high-quality image generator (Instance-Conditioned GAN) to train another mapping from brain response to fine-grained image features using a variational approach (Cortex2Detail). Image reconstructions obtained by our proposed method achieve state-of-the-art semantic fidelity, while yielding good fine-grained similarity with the ground-truth stimulus. Our code is available at: https://github.com/zijin-gu/meshconv-decoding.git.
♻ ☆ Encoder-based Domain Tuning for Fast Personalization of Text-to-Image Models
Text-to-image personalization aims to teach a pre-trained diffusion model to reason about novel, user provided concepts, embedding them into new scenes guided by natural language prompts. However, current personalization approaches struggle with lengthy training times, high storage requirements or loss of identity. To overcome these limitations, we propose an encoder-based domain-tuning approach. Our key insight is that by underfitting on a large set of concepts from a given domain, we can improve generalization and create a model that is more amenable to quickly adding novel concepts from the same domain. Specifically, we employ two components: First, an encoder that takes as an input a single image of a target concept from a given domain, e.g. a specific face, and learns to map it into a word-embedding representing the concept. Second, a set of regularized weight-offsets for the text-to-image model that learn how to effectively ingest additional concepts. Together, these components are used to guide the learning of unseen concepts, allowing us to personalize a model using only a single image and as few as 5 training steps - accelerating personalization from dozens of minutes to seconds, while preserving quality.
comment: Project page at https://tuning-encoder.github.io/
♻ ☆ Improving the Safety of 3D Object Detectors in Autonomous Driving using IoGT and Distance Measures IROS 2023
State-of-the-art object detectors are commonly evaluated based on accuracy metrics such as mean Average Precision (mAP). In this paper, inspired by the fact that mAP is not a direct safety indicator, we propose a straightforward safety metric, especially for 3D object detectors in Autonomous Driving contexts, by combining the Intersection-over-Ground-Truth (IoGT) measure and a distance ratio. Subsequently, we formulate a safety-aware loss function by amending IoGT to commonly used accuracy-oriented loss functions. Our experiments using models from the MMDetection3D library, the nuScenes dataset, and an in-house simulation dataset demonstrate that the object detector trained with our loss function significantly reduces unsafe predictions while staying performant on accuracy and maintaining good stability in the learning process.
comment: 8 pages (IEEE double column format), 8 figures, revised with clearer presentation and resubmitted to IROS 2023
♻ ☆ Automation of reversible steganographic coding with nonlinear discrete optimisation
Authentication mechanisms are at the forefront of defending the world from various types of cybercrime. Steganography can serve as an authentication solution through the use of a digital signature embedded in a carrier object to ensure the integrity of the object and simultaneously lighten the burden of metadata management. Nevertheless, despite being generally imperceptible to human sensory systems, any degree of steganographic distortion might be inadmissible in fidelity-sensitive situations such as forensic science, legal proceedings, medical diagnosis and military reconnaissance. This has led to the development of reversible steganography. A fundamental element of reversible steganography is predictive analytics, for which powerful neural network models have been effectively deployed. Another core element is reversible steganographic coding. Contemporary coding is based primarily on heuristics, which offers a shortcut towards sufficient, but not necessarily optimal, capacity--distortion performance. While attempts have been made to realise automatic coding with neural networks, perfect reversibility is unattainable via such learning machinery. Instead of relying on heuristics and machine learning, we aim to derive optimal coding by means of mathematical optimisation. In this study, we formulate reversible steganographic coding as a nonlinear discrete optimisation problem with a logarithmic capacity constraint and a quadratic distortion objective. Linearisation techniques are developed to enable iterative mixed-integer linear programming. Experimental results validate the near-optimality of the proposed optimisation algorithm when benchmarked against a brute-force method.
♻ ☆ Improving Feature-based Visual Localization by Geometry-Aided Matching
Feature matching is crucial in visual localization, where 2D-3D correspondence plays a major role in determining the accuracy of camera pose. A sufficient number of well-distributed 2D-3D correspondences is essential for accurate pose estimation due to noise. However, existing 2D-3D feature matching methods rely on finding nearest neighbors in the feature space and removing outliers using hand-crafted heuristics, which may lead to potential matches being missed or the correct matches being filtered out. In this work, we propose a novel method called Geometry-Aided Matching (GAM), which incorporates both appearance information and geometric context to address this issue and to improve 2D-3D feature matching. GAM can greatly boost the recall of 2D-3D matches while maintaining high precision. We apply GAM to a new hierarchical visual localization pipeline and show that GAM can effectively improve the robustness and accuracy of localization. Extensive experiments show that GAM can find more real matches than hand-crafted heuristics and learning baselines. Our proposed localization method achieves state-of-the-art results on multiple visual localization datasets. Experiments on Cambridge Landmarks dataset show that our method outperforms the existing state-of-the-art methods and is six times faster than the top-performed method. The source code is available at https://github.com/openxrlab/xrlocalization.
♻ ☆ Masked Jigsaw Puzzle: A Versatile Position Embedding for Vision Transformers CVPR2023
Position Embeddings (PEs), an arguably indispensable component in Vision Transformers (ViTs), have been shown to improve the performance of ViTs on many vision tasks. However, PEs have a potentially high risk of privacy leakage since the spatial information of the input patches is exposed. This caveat naturally raises a series of interesting questions about the impact of PEs on the accuracy, privacy, prediction consistency, etc. To tackle these issues, we propose a Masked Jigsaw Puzzle (MJP) position embedding method. In particular, MJP first shuffles the selected patches via our block-wise random jigsaw puzzle shuffle algorithm, and their corresponding PEs are occluded. Meanwhile, for the non-occluded patches, the PEs remain the original ones but their spatial relation is strengthened via our dense absolute localization regressor. The experimental results reveal that 1) PEs explicitly encode the 2D spatial relationship and lead to severe privacy leakage problems under gradient inversion attack; 2) Training ViTs with the naively shuffled patches can alleviate the problem, but it harms the accuracy; 3) Under a certain shuffle ratio, the proposed MJP not only boosts the performance and robustness on large-scale datasets (i.e., ImageNet-1K and ImageNet-C, -A/O) but also improves the privacy preservation ability under typical gradient attacks by a large margin. The source code and trained models are available at~\url{https://github.com/yhlleo/MJP}.
comment: Accepted to CVPR2023
♻ ☆ Efficient End-to-End Video Question Answering with Pyramidal Multimodal Transformer AAAI 2023
This paper presents a new method for end-to-end Video Question Answering (VideoQA), aside from the current popularity of using large-scale pre-training with huge feature extractors. We achieve this with a pyramidal multimodal transformer (PMT) model, which simply incorporates a learnable word embedding layer, a few convolutional and transformer layers. We use the anisotropic pyramid to fulfill video-language interactions across different spatio-temporal scales. In addition to the canonical pyramid, which includes both bottom-up and top-down pathways with lateral connections, novel strategies are proposed to decompose the visual feature stream into spatial and temporal sub-streams at different scales and implement their interactions with the linguistic semantics while preserving the integrity of local and global semantics. We demonstrate better or on-par performances with high computational efficiency against state-of-the-art methods on five VideoQA benchmarks. Our ablation study shows the scalability of our model that achieves competitive results for text-to-video retrieval by leveraging feature extractors with reusable pre-trained weights, and also the effectiveness of the pyramid.
comment: Accepted by AAAI 2023
♻ ☆ Ev-NeRF: Event Based Neural Radiance Field WACV 2023
We present Ev-NeRF, a Neural Radiance Field derived from event data. While event cameras can measure subtle brightness changes in high frame rates, the measurements in low lighting or extreme motion suffer from significant domain discrepancy with complex noise. As a result, the performance of event-based vision tasks does not transfer to challenging environments, where the event cameras are expected to thrive over normal cameras. We find that the multi-view consistency of NeRF provides a powerful self-supervision signal for eliminating the spurious measurements and extracting the consistent underlying structure despite highly noisy input. Instead of posed images of the original NeRF, the input to Ev-NeRF is the event measurements accompanied by the movements of the sensors. Using the loss function that reflects the measurement model of the sensor, Ev-NeRF creates an integrated neural volume that summarizes the unstructured and sparse data points captured for about 2-4 seconds. The generated neural volume can also produce intensity images from novel views with reasonable depth estimates, which can serve as a high-quality input to various vision-based tasks. Our results show that Ev-NeRF achieves competitive performance for intensity image reconstruction under extreme noise conditions and high-dynamic-range imaging.
comment: Accepted to WACV 2023
♻ ☆ Dual Diffusion Implicit Bridges for Image-to-Image Translation ICLR 2023
Common image-to-image translation methods rely on joint training over data from both source and target domains. The training process requires concurrent access to both datasets, which hinders data separation and privacy protection; and existing models cannot be easily adapted for translation of new domain pairs. We present Dual Diffusion Implicit Bridges (DDIBs), an image translation method based on diffusion models, that circumvents training on domain pairs. Image translation with DDIBs relies on two diffusion models trained independently on each domain, and is a two-step process: DDIBs first obtain latent encodings for source images with the source diffusion model, and then decode such encodings using the target model to construct target images. Both steps are defined via ordinary differential equations (ODEs), thus the process is cycle consistent only up to discretization errors of the ODE solvers. Theoretically, we interpret DDIBs as concatenation of source to latent, and latent to target Schrodinger Bridges, a form of entropy-regularized optimal transport, to explain the efficacy of the method. Experimentally, we apply DDIBs on synthetic and high-resolution image datasets, to demonstrate their utility in a wide variety of translation tasks and their inherent optimal transport properties.
comment: 18 pages, 12 figures, in the Eleventh International Conference on Learning Representations (ICLR 2023)
♻ ☆ YOLOV: Making Still Image Object Detectors Great at Video Object Detection
Video object detection (VID) is challenging because of the high variation of object appearance as well as the diverse deterioration in some frames. On the positive side, the detection in a certain frame of a video, compared with that in a still image, can draw support from other frames. Hence, how to aggregate features across different frames is pivotal to VID problem. Most of existing aggregation algorithms are customized for two-stage detectors. However, these detectors are usually computationally expensive due to their two-stage nature. This work proposes a simple yet effective strategy to address the above concerns, which costs marginal overheads with significant gains in accuracy. Concretely, different from traditional two-stage pipeline, we select important regions after the one-stage detection to avoid processing massive low-quality candidates. Besides, we evaluate the relationship between a target frame and reference frames to guide the aggregation. We conduct extensive experiments and ablation studies to verify the efficacy of our design, and reveal its superiority over other state-of-the-art VID approaches in both effectiveness and efficiency. Our YOLOX-based model can achieve promising performance (\emph{e.g.}, 87.5\% AP50 at over 30 FPS on the ImageNet VID dataset on a single 2080Ti GPU), making it attractive for large-scale or real-time applications. The implementation is simple, we have made the demo codes and models available at \url{https://github.com/YuHengsss/YOLOV}.
♻ ☆ Mixed-Precision Neural Network Quantization via Learned Layer-wise Importance ECCV 2022
The exponentially large discrete search space in mixed-precision quantization (MPQ) makes it hard to determine the optimal bit-width for each layer. Previous works usually resort to iterative search methods on the training set, which consume hundreds or even thousands of GPU-hours. In this study, we reveal that some unique learnable parameters in quantization, namely the scale factors in the quantizer, can serve as importance indicators of a layer, reflecting the contribution of that layer to the final accuracy at certain bit-widths. These importance indicators naturally perceive the numerical transformation during quantization-aware training, which can precisely provide quantization sensitivity metrics of layers. However, a deep network always contains hundreds of such indicators, and training them one by one would lead to an excessive time cost. To overcome this issue, we propose a joint training scheme that can obtain all indicators at once. It considerably speeds up the indicators training process by parallelizing the original sequential training processes. With these learned importance indicators, we formulate the MPQ search problem as a one-time integer linear programming (ILP) problem. That avoids the iterative search and significantly reduces search time without limiting the bit-width search space. For example, MPQ search on ResNet18 with our indicators takes only 0.06 s, which improves time efficiency exponentially compared to iterative search methods. Also, extensive experiments show our approach can achieve SOTA accuracy on ImageNet for far-ranging models with various constraints (e.g., BitOps, compress rate). Code is available on https://github.com/1hunters/LIMPQ.
comment: Published on ECCV 2022, code is available on https://github.com/1hunters/LIMPQ
Machine Learning
☆ CoRTX: Contrastive Framework for Real-time Explanation
Recent advancements in explainable machine learning provide effective and faithful solutions for interpreting model behaviors. However, many explanation methods encounter efficiency issues, which largely limit their deployments in practical scenarios. Real-time explainer (RTX) frameworks have thus been proposed to accelerate the model explanation process by learning a one-feed-forward explainer. Existing RTX frameworks typically build the explainer under the supervised learning paradigm, which requires large amounts of explanation labels as the ground truth. Considering that accurate explanation labels are usually hard to obtain due to constrained computational resources and limited human efforts, effective explainer training is still challenging in practice. In this work, we propose a COntrastive Real-Time eXplanation (CoRTX) framework to learn the explanation-oriented representation and relieve the intensive dependence of explainer training on explanation labels. Specifically, we design a synthetic strategy to select positive and negative instances for the learning of explanation. Theoretical analysis show that our selection strategy can benefit the contrastive learning process on explanation tasks. Experimental results on three real-world datasets further demonstrate the efficiency and efficacy of our proposed CoRTX framework.
☆ Improved Sample Complexity Bounds for Distributionally Robust Reinforcement Learning AISTATS
We consider the problem of learning a control policy that is robust against the parameter mismatches between the training environment and testing environment. We formulate this as a distributionally robust reinforcement learning (DR-RL) problem where the objective is to learn the policy which maximizes the value function against the worst possible stochastic model of the environment in an uncertainty set. We focus on the tabular episodic learning setting where the algorithm has access to a generative model of the nominal (training) environment around which the uncertainty set is defined. We propose the Robust Phased Value Learning (RPVL) algorithm to solve this problem for the uncertainty sets specified by four different divergences: total variation, chi-square, Kullback-Leibler, and Wasserstein. We show that our algorithm achieves $\tilde{\mathcal{O}}(|\mathcal{S}||\mathcal{A}| H^{5})$ sample complexity, which is uniformly better than the existing results by a factor of $|\mathcal{S}|$, where $|\mathcal{S}|$ is number of states, $|\mathcal{A}|$ is the number of actions, and $H$ is the horizon length. We also provide the first-ever sample complexity result for the Wasserstein uncertainty set. Finally, we demonstrate the performance of our algorithm using simulation experiments.
comment: Appeared in The 26th (2023) International Conference on Artificial Intelligence and Statistics (AISTATS)
♻ ☆ Prediction of Gender from Longitudinal MRI data via Deep Learning on Adolescent Data Reveals Unique Patterns Associated with Brain Structure and Change over a Two-year Period
Deep learning algorithms for predicting neuroimaging data have shown considerable promise in various applications. Prior work has demonstrated that deep learning models that take advantage of the data's 3D structure can outperform standard machine learning on several learning tasks. However, most prior research in this area has focused on neuroimaging data from adults. Within the Adolescent Brain and Cognitive Development (ABCD) dataset, a large longitudinal development study, we examine structural MRI data to predict gender and identify gender-related changes in brain structure. Results demonstrate that gender prediction accuracy is exceptionally high (>97%) with training epochs >200 and that this accuracy increases with age. Brain regions identified as the most discriminative in the task under study include predominantly frontal areas and the temporal lobe. When evaluating gender predictive changes specific to a two-year increase in age, a broader set of visual, cingulate, and insular regions are revealed. Our findings show a robust gender-related structural brain change pattern, even over a small age range. This suggests that it might be possible to study how the brain changes during adolescence by looking at how these changes are related to different behavioral and environmental factors.
comment: I submitted the wrong paper
♻ ☆ A Sidecar Separator Can Convert a Single-Talker Speech Recognition System to a Multi-Talker One ICASSP
Although automatic speech recognition (ASR) can perform well in common non-overlapping environments, sustaining performance in multi-talker overlapping speech recognition remains challenging. Recent research revealed that ASR model's encoder captures different levels of information with different layers -- the lower layers tend to have more acoustic information, and the upper layers more linguistic. This inspires us to develop a Sidecar separator to empower a well-trained ASR model for multi-talker scenarios by separating the mixed speech embedding between two suitable layers. We experimented with a wav2vec 2.0-based ASR model with a Sidecar mounted. By freezing the parameters of the original model and training only the Sidecar (8.7 M, 8.4% of all parameters), the proposed approach outperforms the previous state-of-the-art by a large margin for the 2-speaker mixed LibriMix dataset, reaching a word error rate (WER) of 10.36%; and obtains comparable results (7.56%) for LibriSpeechMix dataset when limited training.
comment: Accepted by IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), 2023
♻ ☆ A Machine Learning Case Study for AI-empowered echocardiography of Intensive Care Unit Patients in low- and middle-income countries
We present a Machine Learning (ML) study case to illustrate the challenges of clinical translation for a real-time AI-empowered echocardiography system with data of ICU patients in LMICs. Such ML case study includes data preparation, curation and labelling from 2D Ultrasound videos of 31 ICU patients in LMICs and model selection, validation and deployment of three thinner neural networks to classify apical four-chamber view. Results of the ML heuristics showed the promising implementation, validation and application of thinner networks to classify 4CV with limited datasets. We conclude this work mentioning the need for (a) datasets to improve diversity of demographics, diseases, and (b) the need of further investigations of thinner models to be run and implemented in low-cost hardware to be clinically translated in the ICU in LMICs. The code and other resources to reproduce this work are available at https://github.com/vital-ultrasound/ai-assisted-echocardiography-for-low-resource-countries.
♻ ☆ On the Mathematics of Diffusion Models
This paper gives direct derivations of the differential equations and likelihood formulas of diffusion models assuming only knowledge of Gaussian distributions. A VAE analysis derives both forward and backward stochastic differential equations (SDEs) as well as non-variational integral expressions for likelihood formulas. A score-matching analysis derives the reverse diffusion ordinary differential equation (ODE) and a family of reverse-diffusion SDEs parameterized by noise level. The paper presents the mathematics directly with attributions saved for a final section.
♻ ☆ A Time Series is Worth 64 Words: Long-term Forecasting with Transformers ICLR 2023
We propose an efficient design of Transformer-based models for multivariate time series forecasting and self-supervised representation learning. It is based on two key components: (i) segmentation of time series into subseries-level patches which are served as input tokens to Transformer; (ii) channel-independence where each channel contains a single univariate time series that shares the same embedding and Transformer weights across all the series. Patching design naturally has three-fold benefit: local semantic information is retained in the embedding; computation and memory usage of the attention maps are quadratically reduced given the same look-back window; and the model can attend longer history. Our channel-independent patch time series Transformer (PatchTST) can improve the long-term forecasting accuracy significantly when compared with that of SOTA Transformer-based models. We also apply our model to self-supervised pre-training tasks and attain excellent fine-tuning performance, which outperforms supervised training on large datasets. Transferring of masked pre-trained representation on one dataset to others also produces SOTA forecasting accuracy. Code is available at: https://github.com/yuqinie98/PatchTST.
comment: Accepted by ICLR 2023
Multimedia
☆ Time-frequency Network for Robust Speaker Recognition
The wide deployment of speech-based biometric systems usually demands high-performance speaker recognition algorithms. However, most of the prior works for speaker recognition either process the speech in the frequency domain or time domain, which may produce suboptimal results because both time and frequency domains are important for speaker recognition. In this paper, we attempt to analyze the speech signal in both time and frequency domains and propose the time-frequency network~(TFN) for speaker recognition by extracting and fusing the features in the two domains. Based on the recent advance of deep neural networks, we propose a convolution neural network to encode the raw speech waveform and the frequency spectrum into domain-specific features, which are then fused and transformed into a classification feature space for speaker recognition. Experimental results on the publicly available datasets TIMIT and LibriSpeech show that our framework is effective to combine the information in the two domains and performs better than the state-of-the-art methods for speaker recognition.
comment: 5pages, 3 figures
☆ Heterogeneous Graph Learning for Acoustic Event Classification
Heterogeneous graphs provide a compact, efficient, and scalable way to model data involving multiple disparate modalities. This makes modeling audiovisual data using heterogeneous graphs an attractive option. However, graph structure does not appear naturally in audiovisual data. Graphs for audiovisual data are constructed manually which is both difficult and sub-optimal. In this work, we address this problem by (i) proposing a parametric graph construction strategy for the intra-modal edges, and (ii) learning the crossmodal edges. To this end, we develop a new model, heterogeneous graph crossmodal network (HGCN) that learns the crossmodal edges. Our proposed model can adapt to various spatial and temporal scales owing to its parametric construction, while the learnable crossmodal edges effectively connect the relevant nodes across modalities. Experiments on a large benchmark dataset (AudioSet) show that our model is state-of-the-art (0.53 mean average precision), outperforming transformer-based models and other graph-based models.
comment: arXiv admin note: text overlap with arXiv:2207.07935
☆ Cyber Vaccine for Deepfake Immunity
Deepfakes pose an evolving threat to cybersecurity, which calls for the development of automated countermeasures. While considerable forensic research has been devoted to the detection and localisation of deepfakes, solutions for reversing fake to real are yet to be developed. In this study, we introduce cyber vaccination for conferring immunity to deepfakes. Analogous to biological vaccination that injects antigens to induce immunity prior to infection by an actual pathogen, cyber vaccination simulates deepfakes and performs adversarial training to build a defensive immune system. Aiming at building up attack-agnostic immunity with limited computational resources, we propose to simulate various deepfakes with one single overpowered attack: face masking. The proposed immune system consists of a vaccinator for inducing immunity and a neutraliser for recovering facial content. Experimental evaluations demonstrate effective immunity to face replacement, face reenactment and various types of corruptions.
☆ VTQA: Visual Text Question Answering via Entity Alignment and Cross-Media Reasoning
The ideal form of Visual Question Answering requires understanding, grounding and reasoning in the joint space of vision and language and serves as a proxy for the AI task of scene understanding. However, most existing VQA benchmarks are limited to just picking the answer from a pre-defined set of options and lack attention to text. We present a new challenge with a dataset that contains 23,781 questions based on 10124 image-text pairs. Specifically, the task requires the model to align multimedia representations of the same entity to implement multi-hop reasoning between image and text and finally use natural language to answer the question. The aim of this challenge is to develop and benchmark models that are capable of multimedia entity alignment, multi-step reasoning and open-ended answer generation.
♻ ☆ Automation of reversible steganographic coding with nonlinear discrete optimisation
Authentication mechanisms are at the forefront of defending the world from various types of cybercrime. Steganography can serve as an authentication solution through the use of a digital signature embedded in a carrier object to ensure the integrity of the object and simultaneously lighten the burden of metadata management. Nevertheless, despite being generally imperceptible to human sensory systems, any degree of steganographic distortion might be inadmissible in fidelity-sensitive situations such as forensic science, legal proceedings, medical diagnosis and military reconnaissance. This has led to the development of reversible steganography. A fundamental element of reversible steganography is predictive analytics, for which powerful neural network models have been effectively deployed. Another core element is reversible steganographic coding. Contemporary coding is based primarily on heuristics, which offers a shortcut towards sufficient, but not necessarily optimal, capacity--distortion performance. While attempts have been made to realise automatic coding with neural networks, perfect reversibility is unattainable via such learning machinery. Instead of relying on heuristics and machine learning, we aim to derive optimal coding by means of mathematical optimisation. In this study, we formulate reversible steganographic coding as a nonlinear discrete optimisation problem with a logarithmic capacity constraint and a quadratic distortion objective. Linearisation techniques are developed to enable iterative mixed-integer linear programming. Experimental results validate the near-optimality of the proposed optimisation algorithm when benchmarked against a brute-force method.